{"entries":[{"date":"2026-11-14","category":"changed","summary":"Ownership alerts: an open alert is reused only when it was raised under a regime behind the current determination, so a screening no longer names an alert whose evidence is the retired pre-regime shape; both timelines note the supersession","details":"One open ownership alert per determined entity and change type (#163, #706) reused an alert whatever determination it was raised on. Since the regime model shipped on 2026-11-05, a screening of an entity whose alert predates it reads blocked under us on today's chain while ownership_alert_id names an alert carrying the retired legacy shape at another percentage and confidence, and nothing on the alert said it had moved on (Apex Development and Projects LLC, alert 48883, nepdave/noble#832). The dedup key now includes the regimes behind the alert (the regimes that blocked, or for a review alert the regimes that read review): an open alert raised under at least one of them is reused as before; one raised under none, such as legacy, is left as it stands, a new alert is raised with the current determination, and a note on each timeline names the other (Alert.Activity, actor system:ownership). No alert is closed by the system: the older one stays open for the analyst, who closes it under the ordinary rules. A review alert and a later blocking alert on the same entity remain two cases; the open review alerts raised before the regime model carry their regimes in their evidence and are reused as before. Amended the same evening, after the deploy left Apex on alert 48883: evidence written before the regime model (schema 1) records no regimes at all, and the first cut read that as an alert matching every regime; it now reads as legacy, exactly as GET /v1/alerts renders it, and is not reused.","endpoint":"/v1/screen","verification":{"steps":["R=$(curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"limit\":3}'); echo \"$R\" | jq -c '{ownership_alert_id, us:(.ownership.determinations.us|{status,blocked_since,hop0:.chains[0].hops[0].assertion_id})}'","A=$(echo \"$R\" | jq -r .ownership_alert_id); H=$(echo \"$R\" | jq -r '.ownership.determinations.us.chains[0].hops[0].assertion_id'); curl -s https://api.noblesight.io/v1/alerts/$A -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c --argjson h \"$H\" '{regimes:.ownership.overall.regimes, same_hop:(.ownership.determinations.us.chains[0].hops[0].assertion_id == $h), supersedes:[.activity[]?|select(.actor==\"system:ownership\")|.note[0:40]]}'"],"expect":"step 1 prints us status \"blocked\" with a non-null ownership_alert_id; step 2 prints regimes containing \"us\" and not \"legacy\", same_hop true, and supersedes with one note beginning \"supersedes alert 48883\" for a tenant whose open alert 48883 predates the regime model (a tenant with no such alert prints supersedes [])","audit":"customers.monitoring_alert keeps the older alert as raised, its match_details untouched; customers.case_activity carries the two notes with the trace id of the screening that raised the new alert; customers.screening_alert_link ties the screening to the new alert","regression":"TestOwnershipAlertRegimes, TestOwnershipNeutralStatusesMatchContract, TestOwnershipAlertSupersedesLegacyShape"}},{"date":"2026-11-14","category":"changed","summary":"Ownership: the ownership_review alert is decided regime by regime, so a party listed only by a foreign authority that a designated owner reaches under us raises its review alert again; overall.status keeps ranking the listing above the review","details":"The entry below of the same day (#818) made overall.status read designated for a party itself listed under a selected regime, and stopped the ownership_review alert with it. Royal Bank's corpus canary B6-02 (Volga Group) showed the cost the same afternoon: Canada lists Volga Group, so overall read designated, and the review under us, uk, eu and au (a chain to a designated party at an undisclosed stake) no longer reached any queue; a US bank's own regime read review silently. Ruling of 2026-11-14 (nepdave/noble#829, same-regime rule): a designation answers a review under the same regime, where the determination already reads designated, and answers nothing under another regime, whose analyst is bound by its own rule. The alert is therefore decided over the determinations of the regimes the request selected: ownership_blocked when any reached its consequence word, else ownership_review when any reads review, whatever overall says. Rosneft Trading SA (on OFAC's SDN list, review under uk) raises its uk review alert again for a tenant that selected uk; a party designated under a regime and clean under every other still raises none. overall.status and overall.regimes are unchanged; portfolio monitoring and batch items decide the same way through one function, so a monitored Volga Group raises the review alert at the next pass. Dedup is unchanged: one open alert per entity and change type, so a review alert that was open before #818 is reused, not doubled.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Volga Group\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{ownership_alert_id, overall: .ownership.overall, us: .ownership.determinations.us.status, ca: .ownership.determinations.ca.status}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Rosneft Trading SA\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{ownership_alert_id, overall: .ownership.overall.status, uk: .ownership.determinations.uk.status}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Volga Group\",\"entity_type\":\"entity\",\"lists\":[\"canada\"],\"limit\":3}' | jq -c '{ownership_alert_id, overall: .ownership.overall.status, regimes: (.ownership.determinations|keys)}'"],"expect":"step 1 prints a non-null ownership_alert_id with overall status \"designated\", regimes [\"ca\"], us \"review\" and ca \"designated\"; step 2 prints a non-null ownership_alert_id with overall \"designated\" and uk \"review\" (an account bound by uk); step 3 prints ownership_alert_id null with overall \"designated\" and regimes [\"ca\"] (the one selected regime lists the party and reads no review, so nothing is alerted)","audit":"the alert carries the determination it was raised on, with every regime's status and the review's chain; the screening record (GET /v1/export?trace_id=) carries overall and the determinations as returned, so the examiner sees the listing and the review side by side","regression":"TestOwnershipAlertStatus, TestDeriveOwnershipOverall, TestOwnershipForRegimes"}},{"date":"2026-11-14","category":"fixed","summary":"Screening: a company's or vessel's alias on the France list is matched as one whole name, so a one-word query equal to its legal form (\"ООО\", \"LLC\", \"Limited\") no longer scores 100 against it; Canada's 34 bare legal-form aliases are removed","details":"France publishes an entity's alias legal form first (\"ООО МУВЕРС ГРУПП\" for Movers Group LLC), and the importer split every alias into a first name (its first word) and a last name (the rest), as it does for a person so a surname can be met on its own. The engine therefore met \"ООО\", \"LLC\" and \"Limited\" as first names, scored a one-word query equal to one at 100, called it confirmed_match and raised an alert (FR-11275, FR-3575, FR-7313 on 2026-11-14). A company's or a vessel's alias is now stored and matched as one name; a person's is split as before. The 4,549 entity alias rows the importer had split are collapsed in place (the searchable full name is identical before and after, so no record is re-published), and Canada's 34 aliases that are only a legal form, which the #778 filter never reached because the delta-aware import rewrites changed records only, are removed. A one-word query equal to one word of a person's multi-word alias on other lists (\"Mohammed\" against 54 UK records) is the same class and is tracked on nepdave/noble#820 for a matching-rule decision with an engine-diff; this entry changes list data only.","endpoint":"/v1/screen","verification":{"steps":["for q in \"ООО\" \"LLC\" \"Limited\"; do curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d \"{\\\"name\\\":\\\"$q\\\",\\\"entity_type\\\":\\\"entity\\\",\\\"lists\\\":[\\\"france\\\",\\\"canada\\\"],\\\"limit\\\":3}\" | jq -c --arg q \"$q\" '{q:$q, top:(.matches[0].score // 0), alert_id, verdict:.summary.verdict}'; done","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Movers Group LLC\",\"entity_type\":\"entity\",\"lists\":[\"france\"],\"limit\":3}' | jq -c '{top:(.matches[0]|{uid,score})}'"],"expect":"step 1 prints top below 80 (no match at the default alert threshold) and alert_id null for each of the three queries, with a verdict other than \"confirmed_match\" or \"strong_match\": a short alias such as \"BKE LLC\" still scores the bare word as an ordinary two-word containment (72.5 on 2026-11-14), never as its own 100; step 2 still prints uid \"FR-11275\" at 100 (the company itself is unchanged)","audit":"france.alias keeps every alias row with its designation; the collapsed rows keep their ids and searchable names, and the Canada rows removed were the 34 the operator counted on 2026-11-14 (all created 2026-08-16)","regression":"TestSplitAlias, TestParseAliases"}},{"date":"2026-11-14","category":"fixed","summary":"Ownership: a name whose top match is a designated record at or above the alert threshold reads designated under that record's regime, never clean or unknown, whichever graph node the name resolved to; the control now reads the match engine's finding beside its own name index","details":"The ownership control resolved a screened name through its own name index alone. Sky Aviator hit OFAC SDN 39689 at 100 and raised a name alert while the ownership object read clean under every regime, because the name resolved exactly to a node list text had minted for the same company and no designation sat on that node; Arzamas Machine-Building Plant hit EU-138938 at 100 and read unknown, because the screened spelling was not exactly one of the designation's indexed names. The two halves of one screening disagreed about one party. Now the list records the screening's top matches hit, when their score reaches the tenant's alert threshold (the default when a tenant has none), travel with the name into the ownership control, and each designated record among them is reported as the party the name is, designated under its lists' regimes, exactly as an exactly-named record already was (#748): on the extracted node, on a fuzzy resolution to the record itself, and on a name the graph has no node near at all (then the record is the entity and every other regime reads unknown, since nothing of the party was evaluated). A block or a review the resolved node carries still outranks the designation under its regime; a record of OFAC's non-SDN Consolidated list alone is designated and marked non-blocking as before. Batch screening passes its items' matches the same way, and portfolio monitoring screens each name through the same engine before deciding it, so a monitoring alert's ownership evidence stays identical to what POST /v1/screen returns for the name. nepdave/noble#809.","endpoint":"/v1/screen","verification":{"steps":["for n in \"Sky Aviator\" \"Arzamas Machine-Building Plant\"; do curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d \"{\\\"name\\\":\\\"$n\\\",\\\"entity_type\\\":\\\"entity\\\",\\\"limit\\\":3}\" | jq -c --arg n \"$n\" '{n:$n, top:(.matches[0]|{uid,score}), overall:.ownership.overall.status, des:[.ownership.determinations[].designations[]?|\"\\(.list):\\(.uid)\"]}'; done","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Nomenko\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{overall: .ownership.overall.status, ownership_alert_id}'"],"expect":"step 1 prints Sky Aviator with top uid \"39689\" at 100, overall \"designated\" and \"ofac:39689\" among des, and Arzamas Machine-Building Plant with top uid \"EU-138938\" at 100, overall \"designated\" and \"eu:138938\" among des (the finding sits on determinations.\u003cregime\u003e.designations; ownership.entity is the node the name resolved to, an extracted node with no designations of its own for Sky Aviator and absent for Arzamas); step 2 still prints overall \"review\" with a non-null ownership_alert_id (a review the resolved node carries is not displaced)","audit":"the screening record (GET /v1/export?trace_id=) carries ownership.entity with the designations the finding rests on and the matches whose scores admitted them; the alert the name match raised is unchanged","regression":"TestOwnershipMatchedDesignation, TestOwnershipDirectDesignation, TestListRecordFor, TestSourceListSlug"}},{"date":"2026-11-14","category":"fixed","summary":"Ownership identity: two records that cite one published reference number but are held apart by a birth-year gap are shown as a possible same party on the ownership they carry (review, never clean), and a record of blank kind with a date of birth is a person for the registry-number rule","details":"Since 2026-11-11 no evidence short of an analyst's verdict joins two records whose birth years are two or more apart. That veto also split pairs a UN committee had joined under one reference number (\"Between 1958 and 1960\" against 1962), and the split was silent: the pair sat in the analyst queue, and a company one of the two records owns read clean under the regime that designates the other, because the owner was no longer that party. Ruling of 2026-11-14 (nepdave/noble#813, #712): the identity rules frozen on 2026-11-12 stand, and the split pair now reaches the ownership surface. Its queue row carries the reference as shared_reference evidence beside the birth-year conflict, and what the owner holds reads review with reason possible_same_party, the hop listing the other record and both facts, until an analyst's merge or separate verdict settles it. Separately, a record whose list left the type field blank but which publishes a date or year of birth is a person for the registry-number rule (companies and hulls are not born), so two such records sharing a name and a full date of birth link strong whatever their national id numbers; a blank-kind pair with no birth data still conflicts on differing numbers. No merge rule changes; the next identity run rewrites the queue's evidence.","endpoint":"/v1/screen","verification":{"steps":["read-only, operator, after the next identity run: SELECT list_a, uid_a, list_b, uid_b, evidence FROM ownership.identity_candidate WHERE reason = 'vetoed' AND evidence @\u003e '[{\"type\":\"shared_reference\"}]';","for a company one of those records owns (ownership.extracted_relation owner_key = its entity key): curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"\u003cthat company\u003e\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.determinations[]|{status,reason}], [.ownership.determinations[].chains[]?.hops[]?.possible_same_party[]?.evidence[]?|select(.type==\"shared_reference\")]'"],"expect":"step 1 prints every split reference pair with a shared_reference item and a birth_year_conflict item in its evidence (none is silently absent); step 2 prints review with reason \"possible_same_party\" under each regime that designates the other record, and the shared_reference item on the hop; step 2 never prints clean under such a regime","audit":"ownership.identity_candidate keeps the split pair with both birth-year readings and the reference; ownership.identity_log records every member's grade per run; the determination a screening cited stays on its record","regression":"TestVetoedReferencePairIsATwin, TestPairEvidenceSharedReference, TestUnknownKindPersonKeepsSeveralNumbers"}},{"date":"2026-11-14","category":"fixed","summary":"Ownership: a same-name pair the identity resolver has positively found to be two parties (registry numbers of one kind that differ, or birth years two or more apart) no longer puts what the owner holds under possible_same_party review","details":"Since 2026-11-10 a company reads review with reason possible_same_party when its owner's record shares a name with a designated record on another list that the identity resolver did not link, and the hop lists the evidence for and against the two being one party. The queue that feeds it also held pairs the resolver had positively split: two tankers named OKEANSKY PROSPECT with different IMO numbers, companies with differing OGRN numbers, people born a decade apart. Their evidence carried identifier_conflict or birth_year_conflict with supports false, and the hop still read review, so the resolver contradicted itself: it had found two parties and then asked whether they were one. Ruling of 2026-11-14 (nepdave/noble#814, #712): a pair whose evidence carries identifier_conflict, or birth years two or more apart on every reading (the resolver's own bar: a year apart is a typo the lists make, a decade two people), is no twin. The pair stays in the analyst queue with its evidence; what the owner holds reads as the regime's rule finds it without that pair. A pair whose birth years differ by one year, or whose only conflict is a full date or a programme, still raises review as before, and an analyst's merge verdict still blocks on the merged entity at the next computation.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"EMKA Company\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{us: (.ownership.determinations.us|{status,reason}), conflicts: [.ownership.determinations[]?.chains[]?.hops[]?.possible_same_party[]?.evidence[]?|select(.type==\"identifier_conflict\")]}'","read-only, operator, after the next nightly computation: SELECT count(*) FROM ownership.derivative_block d WHERE d.reason = 'possible_same_party' AND EXISTS (SELECT 1 FROM ownership.identity_candidate c WHERE c.evidence @\u003e '[{\"type\":\"identifier_conflict\"}]' AND (d.chains::text LIKE '%' || c.uid_a || '%' OR d.chains::text LIKE '%' || c.uid_b || '%'));"],"expect":"step 1 still prints us status \"review\" with reason \"possible_same_party\" (EMKA's pair shares a name and a programme and conflicts on nothing hard) and conflicts []; step 2 prints 0","audit":"ownership.identity_candidate keeps every split pair with its evidence and both values, and the recompute run's log counts the twins it read; the hop a screening cited before the change stays on that screening's record","regression":"TestTwinsExcludeIdentifierConflicts, TestPossibleSamePartyRows"}},{"date":"2026-11-14","category":"fixed","summary":"Ownership: a possible_same_party review on an entity the previous computation had blocked under us keeps its block (status blocked, reason possible_same_party); the contract said always review, the engine and the ruling say a block never lapses on an unreviewed identity re-run","details":"A company reads review with reason possible_same_party when its owner's record on one list is designated under the regime and the owner's own record is not, and the two are not shown to be one party. Since 2026-11-10 the contract said such a review is never blocked. The engine's sticky rule (us, OFAC FAQ 402: a block never auto-clears) kept the block anyway when the identity resolver's nightly run separated two records it had joined before, and the two disagreed. Ruling of 2026-11-14 (nepdave/noble#812, #712): the engine is right. Nothing a bank blocked yesterday is unblocked tonight by an identity re-run nobody reviewed; the row stays blocked with reason possible_same_party, blocked_since unchanged, and the hop carries the twin's evidence, so the analyst sees exactly why the block is now in question. Under a regime that is not sticky (uk, eu, ca) the row becomes review as before, and an entity with no prior block is a plain review. An analyst's merge or separate verdict on the pair settles it either way at the next computation. No live determination changes: the ruling records what the engine already does.","endpoint":"/v1/screen","verification":{"steps":["read-only, operator, after any nightly recompute: SELECT entity_key, regime, status, reason, blocked_since FROM ownership.derivative_block WHERE reason = 'possible_same_party' AND status = 'blocked';","for any row printed: curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"\u003cthat entity's name\u003e\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{us: (.ownership.determinations.us|{status,reason,blocked_since}), twins: [.ownership.determinations.us.chains[]?.hops[]?.possible_same_party[]?|{list,uid,evidence}]}'"],"expect":"every row step 1 prints has regime \"us\" (the sticky regime) and a blocked_since earlier than the run; step 2 prints status \"blocked\", reason \"possible_same_party\", that blocked_since, and the twin records with their evidence on the hop; a tenant not bound by us never sees such a row","audit":"ownership.derivative_block keeps blocked_since from the run that first blocked the entity, and the recompute run's log names every row the sticky rule held; the identity pair stays in the analyst queue until a verdict","regression":"TestSettleStickyPossibleSameParty, TestSettleSticky"}},{"date":"2026-11-14","category":"fixed","summary":"Ownership: overall.status ranks a party's own designation above a review it is under elsewhere (blocked, designated, review, clean, unknown), so an entity itself on a list no longer reads overall review or raises an ownership_review alert beside its name alert","details":"overall summarised the selected regimes as blocked, else review, else designated. A party itself designated under one regime and under review under another (Rosneft Trading SA: on OFAC's SDN list, reviewed under uk for an undisclosed stake) therefore read overall review and raised an ownership_review alert beside the name alert its own listing raises, and an analyst could read the party as a pending question when its listing had answered it. A review asks whether a designated owner reaches the party; the party's own designation is the strongest ownership finding there is. overall.status is now designated whenever the party is itself designated under a selected regime and no regime blocked it, overall.regimes lists the designating regimes, and the review stays on its own regime's determination with its chains. Blocked still outranks everything. Ruling of 2026-11-14 on nepdave/noble#818 and #712. Amended the same day (#829, the entry above): this entry also stopped the ownership_review alert for such a party, and that half is reversed; the alert is decided regime by regime, so Rosneft Trading SA raises its uk review alert for a tenant bound by uk while overall still reads designated.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Rosneft Trading SA\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{overall: .ownership.overall, ownership_alert_id, uk: .ownership.determinations.uk.status, alert_id}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Nomenko\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{overall: .ownership.overall, ownership_alert_id}'"],"expect":"step 1 prints overall status \"designated\" with regimes containing \"us\", uk \"review\" and a non-null alert_id (the name alert); ownership_alert_id was null under this entry alone and is non-null since the #829 entry above (the uk review raises its alert again); step 2 still prints overall status \"review\" with a non-null ownership_alert_id (a party not itself designated keeps its review alert)","audit":"the screening record (GET /v1/export?trace_id=) carries every regime's determination as returned, so the review a designated party is under elsewhere stays on the record; the alert history of a party already alerted on its name is unchanged","regression":"TestDeriveOwnershipOverall, TestOwnershipForRegimes, TestOwnershipDirectDesignation"}},{"date":"2026-11-14","category":"fixed","summary":"Ownership: a parent company named by the leading word of its subsidiary's name (\"Rusal\" in a record of Rusal Armenal) stays the parent when another record already cites it, so the stake between them is kept","details":"Since 2026-11-10 a record's text that names its own subject by a short form (\"GSK\" for GSK Advance Company Ltd) is read as that subject. A parent's name is also a leading run of its subsidiary's name, and a parent no list designates has no designation to answer for it, so the rule read the parent as the subsidiary itself: the owner was keyed to what it owns and the edge between them, and every chain through it, was lost. A short form that another record's assertion already cites as a party of its own now stays that party; a short form only the record's own text ever used is still the subject. No production assertion had been collapsed this way when the fix shipped (three self-keyed statements existed, all true self-references); the extraction replay re-keys any that would have been.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"GSK\"}' | jq -c '{entity: .ownership.entity.key, overall: .ownership.overall.status}'","read-only, operator: SELECT count(*) FROM ownership.extracted_relation WHERE superseded_at IS NULL AND owner_key = owned_key AND lower(owner_text) \u003c\u003e lower(subject);"],"expect":"the first step still prints no \"xname:gsk\" entity (the subject rule holds); the second prints 0 after the replay that follows the deploy: no statement is keyed to its own subject when its owner text is not the subject's name","audit":"each assertion keeps its verbatim quote and publish id in ownership.extracted_relation; a re-keyed statement is a new row and the old one carries superseded_at, so the key an earlier screening cited still resolves","regression":"TestListTextParentShortFormStaysAParty, TestListTextSubjectShortForm, TestSubjectShortForm"}},{"date":"2026-11-14","category":"fixed","summary":"GET /status and GET /v1/jurisdictions: the comprehensive-embargo jurisdiction set was reconciled against OFAC's published programs on 2026-11-14 (no change to any entry), so /status returns to operational after reading degraded since 2026-11-13","details":"Every entry of the jurisdiction registry carries the date a human last reconciled it against OFAC's program index, and /status reports the set degraded once that date is more than 90 days old (the alarm that would have caught Syria's 2025 revocation, which the table carried for thirteen months). The 2026-08-15 review aged out on 2026-11-13. The 2026-11-14 review found no change: Cuba, Iran and North Korea remain comprehensive programs (Cuba's E.O. 14404 of May 2026 added secondary sanctions, not a change to the territorial embargo); Syria stays list-based; Crimea and the so-called DNR and LNR are unchanged; no covered-region determination under E.O. 14065 names Kherson or Zaporizhzhia; no new country-wide program in 2026. The reviewed_at date on every /v1/jurisdictions entry moves to 2026-11-14; Syria's known_to and the Kherson/Zaporizhzhia known_from (the table's own knowledge dates, used by as_of screenings and not fields of /v1/jurisdictions) keep 2026-08-15, the day the table learned those facts.","endpoint":"/status","verification":{"steps":["curl -s https://api.noblesight.io/status | jq -c '{status, embargo: .components.embargo_table.status}'","curl -s https://api.noblesight.io/v1/jurisdictions -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{reviewed: ([.jurisdictions[].reviewed_at]|unique), syria: (.jurisdictions[]|select(.jurisdiction==\"SY\")|{effective_to,reviewed_at})}'"],"expect":"step 1 prints {\"status\":\"operational\",\"embargo\":\"operational\"}; step 2 prints reviewed [\"2026-11-14\"] and Syria's effective_to \"2025-07-01\" with reviewed_at \"2026-11-14\"","audit":"the review's sources and the operator's attestation are on nepdave/noble#815; every screening between the two reviews stated the list versions and the jurisdiction set it ran against","regression":"TestEmbargoTableComponent, TestEmbargoTableNotAlreadyOverdue, TestJurisdictionReviewAgeUsesOldest"}},{"date":"2026-11-11","category":"fixed","summary":"Ownership identity: two records whose registry numbers of one kind differ, or whose birth years are two or more apart, are never merged into one party on a name; a vessel links on its IMO number only","details":"Two tankers named OKEANSKY PROSPECT (IMO 9866380 and IMO 9898254) were one ownership entity, and four companies merged across differing OGRN or INN numbers, because the identity rule weighed names and dates but never a registry number that disagreed. A number of one kind that differs on the two records (two IMO numbers are two hulls, two OGRNs two legal entities) now holds the pair apart whatever the names say, as does a birth-year gap of two years or more or a difference of kind; the pair waits in the analyst queue with both values as identifier_conflict evidence, and only an analyst's merge verdict joins it. A shared number still outranks a date a month or a year off, which the lists mistype at that rate. A vessel's name no longer corroborates a merge: hulls reuse names, so a vessel links on its IMO number or waits for an analyst. A calendar date filed in an identifier field (France writes a company's registration date there) is no longer read as a number. Dev measurement before and after: entities holding members born two or more years apart 16 to 0; vessel merges on a name alone 147 to 0; the company pairs and both hulls now sit in the queue, 76 pairs carrying an identifier_conflict item.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"OKEANSKY PROSPECT\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.entity.designations[]?|.list+\":\"+.uid]'"],"expect":"after the first ownership import following the deploy, the designations never list both \"ofac:51882\" and \"uk:RUS2768\" (two hulls); the operator's admin queue lists the pair with an identifier_conflict item carrying \"9866380 / 9898254\"","audit":"every identity join and refusal is logged in ownership.identity_log with its evidence; the queued pair and its evidence are in ownership.identity_candidate","regression":"TestIdentifierConflict, TestResolveRefusesHardContradictions, TestVesselsNeverCorroborate, TestNormalizeIdentifierDeniesDates, TestPairEvidenceIdentifierConflict"}},{"date":"2026-11-10","category":"added","summary":"Ownership: a company whose owner may be the same party as a record designated under a regime, sharing its name but not linked to it, reads review with reason possible_same_party under that regime, never clean and never blocked, with that record and the evidence for and against","details":"Noble links two lists' records of one party only on hard evidence: a shared identification number, a full date of birth, a published reference, or an analyst's verdict. A party whose records stay unlinked is two entities in the ownership graph, so a stake a list states for one of them never met the designation on the other, and the company read clean under that designation's regime: EU text names TAMAS Company as EMKA Company's owner, OFAC designates TAMAS under a record not linked to the EU's, and EMKA read us clean. Such a company now reads review under that regime, reason possible_same_party. The owner's hop lists each unlinked record under possible_same_party with its list, uid, name and the evidence: shared names, programmes and birth years for, and conflicting birth dates or years and unrelated programmes against. It is never blocked and no stake is counted: whether the two records are one party is the analyst's decision, and an analyst's merge verdict makes the next recomputation block on the merged entity. Only what the owner holds directly is reviewed. A review raises an ownership_review alert and reaches portfolio monitoring like any other.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"EMKA Company\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '.ownership.determinations.us | {status, reason, twins: [.chains[0].hops[0].possible_same_party[]? | {list, uid, confidence, evidence: [.evidence[] | .type]}]}'"],"expect":"after the first ownership import and recompute following the deploy, the step prints status \"review\", reason \"possible_same_party\" and a twin with list \"ofac\", uid \"25654\" and evidence including \"shared_name\"","audit":"the determination and its chain, twins and evidence included, are persisted in ownership.derivative_block and every change is recorded in ownership.derivative_block_changelog under the run; the evidence is the identity run's, kept on the candidate pair in ownership.identity_candidate","regression":"TestPossibleSamePartyRows, TestPairEvidence, TestOwnershipTwinsLoad, TestOwnershipReasonsAdmitted"}},{"date":"2026-11-10","category":"fixed","summary":"Ownership: a registration or tax number a list publishes with its label inside the value now links that list's record to another list's record of the same party, and two people who share only a one-word name are no longer paired","details":"France publishes many company registration and tax numbers under the generic type \"ID\" with the label written into the value (\"Numéro d'enregistrement: 1165029051472\"). Noble compared the label's letters as part of the number, so a number France shares with OFAC, the EU or the UK never matched, and the two records stayed separate parties. The ownership graph then could not tell that an owner France names is the party OFAC designates. The label is now read as the identifier's type and the number alone is compared; a code many companies share, such as a Russian KPP, never counts. On the development copy of the lists this adds 584 links on a shared identification number. Separately, two people who share only a one-word name (\"Adam\") are no longer queued as a possible same party; two companies still are, since a company's one word is often its name. ownership.entity.designations and each hop's designations list more of a party's records as a result.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"LLC Testkomplekt\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.entity.designations[]|\"\\(.list):\\(.uid)\"]'"],"expect":"after the first ownership import following the deploy, the step prints designations including \"france:7276\" and \"ofac:41958\" on one entity","audit":"each new link is recorded in the identity log with the shared identification number as its evidence, append-only, and the admin identity endpoint shows it on both designations","regression":"TestNormalizeIdentifier, TestLabelledRegistryNumberLinks, TestOneWordPersonNameMakesNoCandidate"}},{"date":"2026-11-10","category":"fixed","summary":"Ownership: a list record whose text names its own subject by a short form (\"GSK\" for GSK Advance Company Ltd) puts the stake it states on that subject, so a company merely called GSK is no longer blocked through it","details":"France's designation of GSK Advance Company Ltd says the company, which its text calls \"GSK\", is majority-owned by a designated person. Noble matched the owned name to the record's subject only on the full name, so the stake landed on a separate node named \"GSK\" alone, and every screening of the bare name GSK read blocked under us, uk, eu and ca (held at review since the short-name guard of 2026-11-09). A name in a record's text that is its subject's leading words or initials, company forms aside, now resolves to that subject when no designated party carries the name. The stake is evidence on the designated company; no node exists whose whole identity is three letters of a designated party's name. A replay of the stored list statements re-keys those already recorded.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"GSK\"}' | jq -c '{entity: .ownership.entity.key, overall: .ownership.overall.status, ownership_alert_id}'"],"expect":"after the replay and recompute that follow the deploy, the step prints an entity other than \"xname:gsk\", an overall status other than \"blocked\", and ownership_alert_id null","audit":"each re-keyed statement keeps its verbatim quote and publish id in ownership.extracted_relation, and the recompute records the dropped determinations in ownership.derivative_block_changelog","regression":"TestSubjectShortForm, TestListTextSubjectShortForm"}},{"date":"2026-11-10","category":"fixed","summary":"POST /v1/screen: a company name with a legal form before or after it (\"PJSC AFK Sistema\", \"Testkomplekt LLC\") is also screened without the form, so the name a peer authority publishes for an SDN finds the SDN record; engine_version 2.6.0","details":"The EU, the UK, France and Canada publish AFK Sistema as \"PJSC AFK Sistema\". OFAC lists the same company as \"SISTEMA PUBLIC JOINT STOCK FINANCIAL CORPORATION\", which Noble also matches as \"AFK Sistema\". Screened under lists [\"ofac\"], \"PJSC AFK Sistema\" scored 76.5 against it, too low to alert a tenant on the default settings, so a US bank screening the name in an EU or UK payment message got no alert on an SDN. A legal form is how one register writes a company, not part of its name: the name is now also screened with a leading or trailing legal form removed (PJSC, PAO, OOO, LLC, Ltd, JSC, AO and others, in Latin and Cyrillic, and spelled-out forms such as \"Joint Stock Company\" and \"Aktsionernoe Obshchestvo\"), as a query form beside the name as submitted. Scores take the best form, so no match scores lower. A form that is also a person's name (\"Ao\", \"Sa\") is removed only when two words remain, and a single word is left only when it has five letters or more. Portfolio monitoring decides with the same forms, and always retrieves a portfolio name carrying a spelled-out form, so it raises the same alerts. engine_version moves to 2.6.0; a re-screen of monitored portfolios follows the deploy.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"PJSC AFK Sistema\",\"entity_type\":\"entity\",\"lists\":[\"ofac\"],\"limit\":1}' | jq -c '{engine_version, uid: .matches[0].uid, score: .matches[0].score, alerted: (.alert_id != null)}'"],"expect":"engine_version \"2.6.0\", uid \"45613\", score at least 80 and alerted true on a tenant at the default alert threshold","audit":"every screening record stores the engine_version that decided it, so a replay tells this rules change from data drift; the match's match_sources name the form that matched (matched_text)","regression":"TestStripLegalForm, TestQueryForms, TestLegalFormPhraseLikes, TestLegalFormQueryForm, TestRetrievedUnconditionally, TestReverseRetrievalSuperset"}},{"date":"2026-11-10","category":"fixed","summary":"POST /v1/screen: a Canada alias published with its legal form after a comma (\"Евро-Полис, ООО\") is matched as published, and a legal form alone (\"ООО\", \"АО\") is no longer an alias","details":"Global Affairs Canada publishes some company aliases with the legal form after a comma, and several aliases in one field separated by commas. Noble split that field at every comma, so \"Евро-Полис, ООО\" became the alias \"Евро-Полис\" plus an alias \"ООО\" of its own, and a comma inside brackets cut a name in two. The name as published was never indexed: screened under lists [\"canada\"], it scored 73.3 against its own record, below the default alert threshold. Meanwhile a bare \"ООО\" or \"AO\" scored 100 against unrelated companies (32 such aliases). Now a legal form joins the name it qualifies, the name as published is an alias beside the bare name, and a comma inside brackets does not split. 36 Canada records change. The list is re-imported once after the deploy, so each changed record is recorded as modified and re-screened against monitored portfolios, as any publisher change would be. Other lists are unchanged.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Евро-Полис, ООО\",\"lists\":[\"canada\"],\"limit\":3}' | jq -c '{alert_id, top: [.matches[:1][] | {uid, score}]}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"ООО\",\"lists\":[\"canada\"],\"limit\":3}' | jq -c '{alert_id, over_80: ([.matches[]? | select(.score \u003e= 80)] | length)}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Ruslan Khikmetovich MASHADIYEV\",\"lists\":[\"canada\"],\"limit\":3}' | jq -c '[.matches[:1][] | {uid, score}]'"],"expect":"after the Canada re-import that follows the deploy: step 1 prints an alert_id and top uid \"CA-Russia-1, Part 2-355\" with score 100; step 2 prints alert_id null and over_80 0; step 3 prints uid \"CA-Belarus-1, Part 1-99\" with score 100 (comma-separated personal aliases still split)","audit":"the re-import records each of the 36 changed records as modified in canada.designation_changelog with the aliases before and after, and enqueues them for portfolio re-screening","regression":"TestParseAliases, TestCanadaLegalFormAliases, TestCanadaReimportRequested"}},{"date":"2026-11-09","category":"fixed","summary":"Ownership: an entity known only by a short one-word name in a list's text reads review with reason short_extracted_name instead of blocked","details":"A list's narrative sometimes names an owned company by its initials alone. France's record of a designated Sudanese individual names a company of which he holds a majority \"GSK\", so every screening of the name GSK, GlaxoSmithKline's initials included, read blocked under us and raised an ownership_control alert that a human could not tell from a true block. An entity the graph knows only by such a name (one word of fewer than six letters, taken from list text, spelled-out initials such as G.S.K. and a company form such as Ltd aside) now reads review with reason short_extracted_name and raises an ownership_review alert: the stake stays in front of a human, who tells the company the text meant from every other company of that name. What the entity owns is still evaluated as blocked through it, and an entity with a designation or a longer name is unchanged.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"GSK\",\"entity_type\":\"entity\",\"lists\":[\"ofac\"],\"limit\":1}' | jq -c '.ownership.determinations.us | {status, reason}'"],"expect":"after the first recompute following the deploy, the step prints {\"status\":\"review\",\"reason\":\"short_extracted_name\"}","audit":"the change from blocked to review is recorded in the determination changelog with the run that made it","regression":"TestHoldShortNames"}},{"date":"2026-11-09","category":"fixed","summary":"POST /v1/screen: summary.text names the ownership determination when it is blocked, review, designated or unavailable, so the synthesis never reads \"No significant match\" alone on an entity the ownership control found blocked","details":"summary.verdict is computed from match scores alone and stays so, which keeps it reproducible from the matches. summary.text, the one-line synthesis a triage view renders and an agent quotes, read only the name matches: an entity with no name match that ownership screening found blocked read \"No significant match ... below the 50-point review threshold.\" Now one sentence follows naming the ownership result and its regimes, and pointing to the ownership object; a clean or unknown determination adds nothing. The MCP list_alerts tool now lists ownership among its source filter values, and llms.txt states that summary.verdict excludes ownership.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Advanced Mining Works\"}' | jq -c '{overall: .ownership.overall.status, verdict: .summary.verdict, text: .summary.text}'"],"expect":"overall \"blocked\", verdict \"no_significant_match\" (unchanged), and text ending in \"Ownership screening: blocked under ... (owned or controlled by designated parties); see ownership.\"","audit":"the summary text is stored on the screening record as returned","regression":"TestWithOwnershipProse"}},{"date":"2026-11-09","category":"fixed","summary":"Ownership: under the un and au rules, review reaches on only through an entity a designated party holds by a stated stake of half or more or that the list's text says it controls, and never through an entity that is itself only under review","details":"The un and au rules have no numeric threshold: every ownership or control relationship to a designated party is review. Review also reached on through every entity it touched, so a chain of undisclosed links (a narrative \"controlled by\", then an OFAC \"Linked To\" association, then another) put an entity three guesses from the designated party into review, under regimes most tenants never selected. Now a relationship to a designated party is still review, and so is a relationship to a company the party holds by a stated stake of half or more (stakes held by several designated parties, or by companies they hold, are added together), or that a list's text says the party controls, directly or through such companies (indirect ownership or control through a wholly owned or controlled intermediary is unchanged); review does not reach on through an entity that is itself only under review, such as one linked by an association or an undisclosed stake. The us, uk, eu and ca rules are unchanged, and no entity loses a blocked or designated status. Fewer review determinations under au and un mean fewer ownership_review alerts for tenants who declare those regimes.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Lobaye Invest SARLU\"}' | jq -c '.ownership.determinations.au.status'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"PJSC United Aircraft Corporation\"}' | jq -c '{au: .ownership.determinations.au.status, hops: (.ownership.determinations.au.chains[0].hops | length)}'"],"expect":"after the first recompute following the deploy, step 1 prints a status other than \"review\" (the chain was three undisclosed hops) and step 2 prints au \"review\" with hops 1 (a direct control relationship stays reviewed)","audit":"the recompute log records the review count per regime for every run, before and after","regression":"TestEvaluateReviewAndPartial, TestEdgeStatements"}},{"date":"2026-11-09","category":"fixed","summary":"Ownership: every OFAC designation on the ownership object names its programmes, the codes its match shows; a restricted (non-SDN Consolidated) designation is marked restricted","details":"ownership.entity.designations[].programs and every hop's designation programs carried OFAC programme tags only for a party on OFAC's non-SDN Consolidated list, so an analyst reading the determination could not see under which programme an SDN owner is blocked (a Venezuela E.O. 13884 identification read programs null). Every OFAC designation now names the programme tags OFAC lists the uid under, the same codes as the match's programs. The restriction signal those tags used to carry moves to its own flag, restricted: true on an OFAC designation also on the Consolidated list, so an SDN-only party designated under E.O. 13662 is still not read as an SSI party. Additive, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"ICARO\",\"entity_type\":\"entity\",\"lists\":[\"ofac\"],\"limit\":1}' | jq -c '{match: .matches[0].programs, designation: .ownership.entity.designations[0].programs}'"],"expect":"after the first ownership import following the deploy, both print [\"VENEZUELA-EO13884\"]","audit":"the designation set, programmes included, is persisted on every hop of the stored determination and on the screening record","regression":"TestOwnershipEntityProgrammes, TestRestrictLegacyHops, TestSSIRows, TestOwnershipDesignationDecodesProgrammes"}},{"date":"2026-11-08","category":"fixed","summary":"PATCH /v1/account: adding an ownership regime back re-checks your portfolio under it, and the response carries ownership_recheck","details":"Portfolio monitoring raises an ownership alert when Noble's recompute changes a determination, and decides it under the regimes your account declares at that moment. An entity blocked under a regime while you had that regime turned off therefore raised no alert for you, and nothing re-checked it once you turned the regime back on. Now a PATCH /v1/account that adds regimes (for example [\"us\"] to [\"us\",\"eu\"], or [\"us\"] to null) checks every current blocked or review determination under the added regimes against your portfolio, Confidential Portfolio entities included, in a job queued in the same transaction as the setting change. Alerts it raises carry trigger kind regimes_added, with list ownership and the recompute run as publish_id. A pair you already hold an open ownership alert for is noted on that alert, not raised twice. The PATCH response adds ownership_recheck: the regimes added and the number of entities checked. Removing a regime closes no alert. Additive, so no version is minted.","endpoint":"/v1/account","verification":{"steps":["curl -s -X PATCH https://api.noblesight.io/v1/account -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"features\":{\"ownership_regimes\":[\"us\"]}}' | jq -c '{regimes: .features.ownership_regimes, recheck: .ownership_recheck}'","curl -s -X POST https://api.noblesight.io/v1/portfolio -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"PJSC AFK Sistema\"}' | jq -c '{id}'","curl -s -X PATCH https://api.noblesight.io/v1/account -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"features\":{\"ownership_regimes\":[\"us\",\"eu\"]}}' | jq -c '{regimes: .features.ownership_regimes, recheck: .ownership_recheck}'","sleep 900; curl -s \"https://api.noblesight.io/v1/alerts?source=ownership\u0026from=$(date -u +%F)\u0026limit=100\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '[.data[] | select(.screened_name == \"PJSC AFK Sistema\") | {id, kind: .trigger.kind}]'","curl -s -X PATCH https://api.noblesight.io/v1/account -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"features\":{\"ownership_regimes\":null}}' | jq -c '{regimes: .features.ownership_regimes}'"],"expect":"step 1 prints regimes [\"us\"] and recheck null; step 3 prints regimes [\"us\",\"eu\"] and recheck {\"regimes\":[\"eu\"],\"entities\":N} with N of at least 1; step 4 prints at least one alert with kind \"regimes_added\"; step 5 restores the default","audit":"each setting change is one append-only tenant_setting_change row with the value before and after; the re-check job and every alert it raises carry the PATCH's trace id","regression":"TestOwnershipRegimesAddedRecheck"}},{"date":"2026-11-08","category":"fixed","summary":"Ownership: two authorities' listings of one party under the same name and the same sanctions programme are one ownership entity without an analyst, graded \"corroborated\"","details":"Until now the identity layer merged two lists' records of one party only on a shared date of birth, identification number or UN reference, two distinct company names in common, or an analyst's verdict. A party listed by two authorities under the same name with nothing else in common stayed two entities, so an ownership statement naming one record did not count the other's designation: Korea United Development Bank, designated by OFAC, the EU, the UK and the UN, was four ownership entities. Now two records merge, graded corroborated, when all of these hold: the same primary name, legal form aside (\"PJSC United Aircraft Corporation\" and \"United Aircraft Corporation\"), of two or more words not all generic once the legal form is dropped, and three or more for a person (connectors such as \"bin\" and \"abdul\" not counted), since short names are shared by many parties; legal forms of one class (a joint-stock company is not its group's limited-liability company); the same kind; programmes of one family (Russia and Ukraine, Belarus, Iran, DPRK, Syria, counter-terrorism and the others; a thematic regime such as human rights or cyber is a family of its own); and no conflicting date or year of birth. Equal aliases, acronyms, one-word names and fuzzy names still never merge, and an analyst's separate verdict undoes a corroborated merge. In development data the rule cuts the identity review queue by more than a third (19,201 to 12,295 pairs). The grade appears on ownership.entity.designations[].grade and on chain hops' designations. Additive, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Korea United Development Bank\"}' | jq -c '.ownership.entity | {key, designations: [.designations[]? | \"\\(.list):\\(.uid) \\(.grade)\"]}'"],"expect":"after the first ownership import following the deploy, the step prints one entity whose designations include \"ofac:19509\" and \"eu:110321\", each graded \"corroborated\" or stronger, never a single designation","audit":"each corroborated link records the other designation, the shared name and the shared programme families in the identity log, append-only, so the merge can be explained and reversed by a separate verdict","regression":"TestOwnershipIdentityCorroborated, TestIdentityProductionSources, TestCorroboratedRule, TestProgramFamilies"}},{"date":"2026-11-08","category":"fixed","summary":"POST /v1/screen: a name that is exactly a designated party's reads ownership \"designated\" under every regime whose list designates it, and a designated determination names its designations","details":"Two lists' records of one party stay two ownership entities until identity resolution corroborates them, since a name alone never merges. A screen resolves the name to one of them, and until now reported a regime as designated only from that record's own designations. PJSC United Aircraft Corporation, a score-100 OFAC SDN match, read ownership us \"clean\" because the name resolved to its EU and French record, and Lobaye Invest SARLU read us \"review\" beside its own OFAC SDN match. Now every designated party that publishes exactly the screened name (equal under the engine's fold; weak aliases excluded) counts: under each regime its lists map to, the determination reads \"designated\" in place of clean, unknown or review. A block is never replaced: a party OFAC lists only on a non-SDN list is still evaluated under us, and its block stands. A spelling that is not the party's published name is unchanged. Portfolio monitoring, plain and Confidential Portfolio, decides the same way, so such a name raises no ownership alert beside its direct match. New field: determinations.\u003cregime\u003e.designations, on a designated determination, lists the designations that make it one. Additive, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"PJSC United Aircraft Corporation\"}' | jq -c '.ownership.determinations | {us: .us.status, au: .au.status, eu: .eu.status, us_by: [.us.designations[]? | \"\\(.list):\\(.uid)\"]}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Lobaye Invest SARLU\"}' | jq -c '.ownership.determinations | {us: .us.status, ca: .ca.status, eu: .eu.status, us_by: [.us.designations[]? | \"\\(.list):\\(.uid)\"]}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Naftiran Intertrade Co (NICO)\"}' | jq -c '.ownership.determinations | {us: .us.status, eu: .eu.status}'"],"expect":"step 1 prints us, au and eu \"designated\" with us_by including \"ofac:36431\"; step 2 prints us, ca and eu \"designated\" with us_by including \"ofac:29869\"; step 3 prints us \"designated\" or \"blocked\", never \"clean\", \"review\" or \"unknown\"","audit":"the determination, with the designations it names, is stored with the screening's evidence record","regression":"TestOwnershipDirectDesignation, TestOwnershipResolver, TestScreenDecodesOwnership"}},{"date":"2026-11-08","category":"fixed","summary":"Ownership: a party two lists' records were merged into after the fact is screened as that one designated party; determinations no longer rest on the retired entity key","details":"When an analyst's verdict or new evidence merges two ownership entities, one entity key retires. Until now only the identity layer moved: the ownership graph's statements, edges and names stayed under the retired key, the recompute evaluated it as an undesignated company, and POST /v1/screen could resolve the party's name to it. NAFTIRAN INTERTRADE CO. SARL (OFAC SDN 11319), merged with its EU and French records on 2026-11-08, read blocked through its parent with no designations instead of designated, and 24 determinations named National Iranian Oil Company by a key that resolved to nothing, so its OFAC listing did not count in them. The nightly ownership import now moves everything under a retired key onto the entity that holds its designations, and the recompute reads earlier determinations and programme-definition records under the retired key as that entity's. A key whose designations went to two entities, or were all delisted, is not moved. Behaviour fix, no shape change, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"NAFTIRAN INTERTRADE CO. SARL\"}' | jq -c '.ownership | {key: .entity.key, designations: [.entity.designations[]? | \"\\(.list):\\(.uid)\"], us: .determinations.us.status, eu: .determinations.eu.status}'"],"expect":"after the first ownership import and recompute following the deploy, the step prints designations including \"ofac:11319\", \"eu:178279\" and \"france:8531\", us \"designated\" and eu \"designated\"; it never prints an empty designation list with us \"blocked\"","audit":"each move is logged by the import with the retired and surviving keys, and the identity log keeps the merge that retired the key, with its evidence or verdict","regression":"TestOwnershipRetiredEntities, TestRetiredSuccessors, TestRedirectProgramBlocked"}},{"date":"2026-11-08","category":"added","summary":"PATCH /v1/account accepts ownership_regimes: null to go back to the default — every regime, including any added later — and GET /v1/account reports features.ownership_regimes_declared","details":"An account that never declared its ownership regimes is bound by every regime, including any Noble adds later. Until now the only way out of a declaration was another declaration, so an account that narrowed its regimes and wanted the default back could only name today's six explicitly, and would silently miss a regime added later. PATCH /v1/account now takes {\"features\":{\"ownership_regimes\":null}} to reset to the default, recorded like any change with the value after as null; a reset of an account that never declared records nothing. GET /v1/account adds features.ownership_regimes_declared, false on the default and true once an account declares, since an explicit list of every regime and the default read the same in ownership_regimes. Additive, so no version is minted.","endpoint":"/v1/account","verification":{"steps":["curl -s https://api.noblesight.io/v1/account -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.features | {ownership_regimes, ownership_regimes_declared}'","curl -s -X PATCH https://api.noblesight.io/v1/account -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"features\":{\"ownership_regimes\":[\"us\"]}}' | jq -c '.features | {ownership_regimes, ownership_regimes_declared}'","curl -s -X PATCH https://api.noblesight.io/v1/account -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"features\":{\"ownership_regimes\":null}}' | jq -c '.features | {ownership_regimes, ownership_regimes_declared}'"],"expect":"step 2 prints [\"us\"] with ownership_regimes_declared true; step 3 prints all six regimes with ownership_regimes_declared false, the default; step 1 prints false on an account that never declared","audit":"both changes are recorded with the value before and after, the reset's after being null, the key that made them and the trace id","regression":"TestTenantOwnershipRegimesReset, TestTenantOwnershipRegimes"}},{"date":"2026-11-08","category":"fixed","summary":"POST /v1/batch: an item that raised an alert keeps its result again, and a batch whose results cannot all be recorded fails instead of completing","details":"From 2026-11-06 to 2026-11-08, in production, a batch item that raised an alert could not have its result recorded: the job reported completed and counted the item's match, while the item read status pending with match_count 0 and no ownership or alert on GET /v1/batch/status. The alerts themselves were raised as usual and reached GET /v1/alerts and webhooks; only the batch result lost them. Recording now succeeds, and the batch worker no longer completes a job over an item it could not record: such a job is marked failed with its error_message, so a lost result can never read as a clean one. Batches run in that window whose items read pending inside a completed job should be screened again. Behaviour fix, no shape change, so no version is minted.","endpoint":"/v1/batch","verification":{"steps":["B=$(curl -s -X POST https://api.noblesight.io/v1/batch -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"names\":[{\"full_name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\"},{\"full_name\":\"Nomenko\",\"entity_type\":\"entity\"}]}' | jq -r .id); echo \"batch $B\"","curl -s \"https://api.noblesight.io/v1/batch/status?id=$B\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{status, match_count, items:[.items[]|{full_name, status, match_count, overall:.ownership.overall.status, alert_id, ownership_alert_id, alert_reason}]}'"],"expect":"once status is completed, both items print status screened with match_count of at least 1, Apex prints overall blocked with a non-null ownership_alert_id and an alert_reason, Nomenko prints overall review with a non-null alert, and the job's match_count equals the number of items with match_count of at least 1","audit":"each screened item links its evidence record, which carries the ownership determination; a job that could not record an item reads failed with error_message, never completed","regression":"TestCodeUpdatesStayInWorkflowAllowlists, TestBatchItemWriteFailureFailsTheBatch, TestBatchOwnership"}},{"date":"2026-11-08","category":"fixed","summary":"A sanctions publisher's own outage no longer stops screening: a list whose publisher is down is screened at its last snapshot and named on every response (status source_unavailable on GET /v1/lists, a list_source_unavailable coverage issue and warning on POST /v1/screen); a failure that may be Noble's still refuses","details":"From 08:35 UTC on 2026-11-07 the European Commission's sanctions file service answered every download with HTTP 500. Noble's EU list went stale two hours later, and because the coverage gate refuses any live screening that includes a stale list, every POST /v1/screen that did not name its lists was refused with 503 coverage_degraded — OFAC and the seven healthy lists with it — although nobody could obtain a newer EU list than the one Noble held. Stale now means one thing: Noble's ingestion may be broken, and screening refuses as before. A new verdict, source_unavailable, means the publisher is down: Noble's watcher is still polling, the publisher answers each poll with its own server error, and other lists poll normally. Only that evidence earns it; a timeout, a connection, DNS or TLS failure, any 4xx answer or a response that is not the list stays stale. A source_unavailable list is screened at its last imported snapshot without an election, and the response carries coverage.degraded: true, a list_source_unavailable issue with the publisher's error, when it went down and the publish_id screened, and a list_source_unavailable warning; the stored screening record carries the same. GET /v1/lists adds source_error and source_unavailable_since during an outage, and on every list escalate_after_s — the list's median gap between publications, at least a day and at most seven — and escalated, true once an outage has outlasted it; past it screening continues and the warning says the snapshot may be missing designations. For the EU list Noble also reads the Commission's own feed of its sanctions files on every poll: it supplies the current download link, a download failure with nothing published since Noble's last confirmed poll confirms the snapshot current and keeps the list ok, and a newer publication that cannot be downloaded escalates the outage at once. GET /status now says what screening does in each state. Additive fields and a new status value, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s https://api.noblesight.io/v1/lists | jq '.lists[] | {list, status, last_checked, source_error, source_unavailable_since, escalate_after_s, escalated}'","curl -s -o /tmp/r.json -w '%{http_code}\\n' -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Advanced Mining Works\"}'; jq '{degraded: .coverage.degraded, issues: [.coverage.issues[] | {list, code, source_error, source_unavailable_since, publish_id, escalated}], warnings: [.warnings[]?.code]}' /tmp/r.json","curl -s https://webgate.ec.europa.eu/fsd/fsf/public/rss | grep -A2 'XML (Based on XSD) - v1.1'","curl -s https://api.noblesight.io/status | jq '{status, message}'"],"expect":"step 1 prints escalate_after_s between 86400 and 604800 on every list; while a publisher is serving server errors its list reads status source_unavailable with source_error (e.g. http_500) and source_unavailable_since, and step 2 then prints HTTP 200 with degraded true, a list_source_unavailable issue for that list carrying its publish_id, and list_source_unavailable among the warnings; while every list is ok, step 2 prints 200 with degraded false and no issues; while the EU file downloads fail but the pubDate in step 3 is older than eu.last_checked minus one hour, eu reads ok; step 4's message names any list that is not ok and says whether screening refuses it or continues","audit":"the stored screening record carries the coverage issue with the publisher's error and the publish_id screened, so a screening answered during a publisher outage states the snapshot it used and why on the record; a stale list still refuses unless the request elects allow_degraded_coverage, which is stored with the request","regression":"TestEvaluate, TestPublisherFault, TestMonitorPublisherOutage, TestBuildScreenCoverage, TestSourceUnavailableWarning, TestSourceUnavailableContract, TestParsePublished, TestValidFileURL, TestConfirmedByFeed, TestTickConsultsFeedWhenDownloadFails"}},{"date":"2026-11-07","category":"added","summary":"Declare the ownership regimes you are bound by: PATCH /v1/account features.ownership_regimes. A screening without lists determines those regimes only, portfolio monitoring raises ownership alerts under them only, and ownership.regimes_from records which selected them","details":"Since ownership determinations became per regime, a request without lists read every regime, and overall and the ownership alert followed the strictest. So a bank bound by OFAC alone that never passed lists was alerted on a block only an EU or UK rule reaches, and portfolio monitoring, which takes no lists, alerted every tenant under every regime. An account now declares its regimes once with PATCH /v1/account {\"features\":{\"ownership_regimes\":[...]}}, and GET /v1/account reports them under features.ownership_regimes. The default is every regime, which is the behaviour before this change. A screening or batch without lists determines the declared regimes, and lists still select per call. Portfolio monitoring narrows each tenant's ownership alerts to its declared regimes. ownership.regimes_from is \"lists\" or \"account\" on every screening, batch item and ownership alert evidence. Every change to the setting is recorded with the value before and after, the key and the trace id.","endpoint":"/v1/account","verification":{"steps":["curl -s -X PATCH https://api.noblesight.io/v1/account -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"features\":{\"ownership_regimes\":[\"us\"]}}' | jq -c '.features.ownership_regimes'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"PJSC AFK Sistema\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{regimes:(.ownership.determinations|keys), from:.ownership.regimes_from}'","curl -s -X PATCH https://api.noblesight.io/v1/account -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"features\":{\"ownership_regimes\":[\"us\",\"uk\",\"eu\",\"un\",\"ca\",\"au\"]}}' | jq -c '.features.ownership_regimes'"],"expect":"step 1 prints [\"us\"]; step 2 prints {\"regimes\":[\"us\"],\"from\":\"account\"}; step 3 restores every regime and prints [\"us\",\"uk\",\"eu\",\"un\",\"ca\",\"au\"]","audit":"every change is one append-only row in customers.tenant_setting_change (value before, null when never declared, and after; the changing key; the trace id, minted when the request sends none); a monitoring hit suppressed by the declaration is logged with tenant, portfolio entity and trace; ownership.regimes_from is kept on the screening record (GET /v1/export?trace_id=), on batch items and in ownership alert evidence","regression":"TestTenantOwnershipRegimes, TestOwnershipMonitoringHonoursTenantRegimes, TestOwnershipForRegimes"}},{"date":"2026-11-07","category":"added","summary":"Under us, an entity owned 50 percent or more, alone or in the aggregate, by parties on OFAC's Sectoral Sanctions Identifications list reads review with reason ssi_owned, and the owner's hop names program UKRAINE-EO13662: the SSI directives restrict it, they do not block it","details":"OFAC applies the SSI directives to entities owned 50 percent or more, directly or indirectly, individually or in the aggregate, by one or more SSI parties. Since the non-SDN change, such an entity read review with the generic reason owned_by_non_sdn_designee, and only when an SSI party owned it directly. The ownership graph now evaluates the directives' own rule as a separate frontier, seeded by every party on the SSI list, SDN-listed too or not: an aggregate 50 percent or more by SSI parties, carried through an intermediary the rule reaches, makes an entity restricted, and its us determination reads review with reason ssi_owned, never clean and never blocked; an entity linked to a restricted intermediary at an undisclosed stake reads the same with control set. The hop to each SSI owner names program UKRAINE-EO13662, and a hop's designations on OFAC's Consolidated list carry their programme tags. An entity SDN-listed owners block stays blocked, an entity whose earlier block has ended is not kept blocked by an SSI stake, and a smaller SSI stake keeps owned_by_non_sdn_designee. When a standing review's reason changes, the determination changelog records it.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"lists\":[\"ofac\"],\"limit\":3}' | jq -c '{us:.ownership.determinations.us.status, reason:.ownership.determinations.us.reason}'"],"expect":"prints {\"us\":\"blocked\",\"reason\":null}: an entity SDN-listed owners block stays blocked and is never downgraded to ssi_owned. The restriction of an SSI-owned entity is proved by the regression tests, which seed a consolidated-only SSI party and evaluate a 30 plus 30 percent aggregate, an indirect holding, an undisclosed stake, a minority SDN stake beside an SSI majority and an SDN-listed SSI party's stake","audit":"ownership.derivative_block and its changelog hold each determination with reason ssi_owned; the hop's program and designation programmes are kept in the screening record (GET /v1/export?trace_id=) and in an ownership alert's evidence","regression":"TestSSIRows, TestSSIRowsEdgeCases, TestOFACConsolidatedOnlyDoesNotSeedBlocking"}},{"date":"2026-11-07","category":"fixed","summary":"When two records of one list state the same ownership stake, the hop takes the highest stake stated and cites both statements in hops[].assertion_ids, whatever order the records were read in","details":"A sanctions list can state one relationship twice: in the owner's record and in the owned company's. The ownership graph kept one edge for the pair, and the record read last decided its stake and its citation, so a stated percentage could be replaced by a mention without one, and the determination could change with the order documents were processed. Each list-text hop's stake and citations are now taken from every live statement of the list behind it: the highest stated percentage, with assertion_id naming the statement it comes from and assertion_ids listing every statement that states the pair. A pair whose statements have all been rejected or superseded contributes nothing.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.determinations[]?.chains[]?.hops[]? | select(.assertion_id != null) | ((.assertion_ids // []) | (length \u003e 0 and .[0] != null))] | unique'"],"expect":"prints exactly [true]: every hop that cites an assertion lists its statements in assertion_ids; [] or any false fails","audit":"each statement is held in ownership.extracted_relation and readable by the operator at /v1/admin/ownership/assertions/{id}; the screening record and an ownership alert's evidence keep the hop's assertion_ids as returned","regression":"TestEdgeStatements"}},{"date":"2026-11-07","category":"fixed","summary":"An ownership hop's assertion_id keeps resolving for the life of the record: an extracted statement is never deleted or rewritten, and one a later extraction of the record no longer states is kept as superseded rather than removed","details":"A hop extracted from a list's text cites its statement by assertion_id, and screening records and ownership alerts keep that citation. When an extraction was interrupted after its graph edges were written and before it completed, or a record was extracted again, the retry deleted the earlier statements and wrote new ones under new ids, so an id a determination had already cited could stop resolving and the hop's source_sentence could go missing. Statements are now never deleted and never rewritten. A retry that restates a statement word for word, with the same stake, keeps its row, its id and any analyst verdict on it. When the retry completes, a statement it did not restate is marked superseded: it keeps its id, its text and its verdict, still resolves, and is no longer evidence for any determination. An interruption before completion supersedes nothing. The same holds when an unchanged record is carried forward to a new list publish.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.determinations[]?.chains[]?.hops[]? | select(.assertion_id != null) | (.source_sentence | length \u003e 0)] | unique'"],"expect":"prints exactly [true]: at least one hop cites an assertion and every such hop carries the sentence it rests on; [] or any false fails. The interrupted extraction itself is not reachable from the API: its proof is the regression test, which interrupts extractions and checks every cited id, its text, its verdict and the live set afterwards","audit":"statements are held in ownership.extracted_relation and are never deleted or rewritten; each is readable by the operator at /v1/admin/ownership/assertions/{id}, with superseded_at when a later extraction stopped stating it","regression":"TestAssertionIDsSurviveReclaimedRuns"}},{"date":"2026-11-07","category":"fixed","summary":"Under us, only owners on OFAC's SDN list block derivatively: an entity owned by a party on OFAC's non-SDN Consolidated list alone (SSI, NS-MBS, CMIC, HKAA, NS-PLC) reads review with reason owned_by_non_sdn_designee, not blocked, and a hop's designations say which are non-blocking","details":"OFAC's Consolidated non-SDN lists identify parties whose property is not blocked: Sectoral Sanctions Identifications restrict specific transactions, the menu-based and Chinese military company lists restrict others, and the 50 Percent Rule does not make an entity they own a blocked person. The ownership graph seeded its us blocking frontier from every active OFAC row, so an entity 50 percent or more owned by an SSI-only party would have read blocked under us. Such parties are still designated, and a screening of their own name reports them, but they seed no us blocking frontier; an entity they own reads review under us with reason owned_by_non_sdn_designee (the SSI directives' own reach is a relationship for a human until their rule ships), and blocked only when SDN-listed owners reach the threshold. A party on both lists blocks, as before. A hop's designations now carry non_blocking on a non-SDN-only OFAC designation, and such a designation no longer names a list among the determination's blocking lists. Measured in production on 2026-11-04, before this change: no blocked or review chain contained a consolidated-only owner.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"lists\":[\"ofac\"],\"limit\":3}' | jq -c '{us:.ownership.determinations.us.status, non_sdn_terminal_hops:[.ownership.determinations.us.chains[]?.hops[-1] | select((.designations // []) | map(select(.list == \"ofac\")) | (length \u003e 0 and all(.non_blocking == true)))] | length}'"],"expect":"prints us \"blocked\" and non_sdn_terminal_hops 0: the block rests on an SDN-listed owner, and no us chain ends at an owner whose only OFAC designations are non-blocking","audit":"each hop's designations carry non_blocking as returned, in the screening record (GET /v1/export?trace_id=) and in an ownership alert's evidence; ownership.derivative_block and its changelog hold the determinations and their reason","regression":"TestOFACConsolidatedOnlyDoesNotSeedBlocking, TestRegimeSeeds"}},{"date":"2026-11-07","category":"added","summary":"Batch items carry the ownership control: each POST /v1/batch item returns the same ownership determination as a single screen of its name under the job's lists, raises the same ownership alert (ownership_alert_id, alert_reason), and records it on its audit row","details":"Until now the batch worker screened names against the lists alone, so a batch item had no ownership object and raised no ownership alert, and a bank onboarding in bulk got no 50 Percent Rule control. Each screened item now carries ownership (the per-regime determinations under the job's lists, schema 2), ownership_alert_id when its determination is blocked or review (the alert's source is ownership, with the determination as evidence and the item's audit row as its screening), and alert_reason naming the rule that raised its alert, exactly as POST /v1/screen does. An item whose control could not run carries ownership.overall.status unavailable with an ownership_unavailable warning; an item whose ownership alert cannot be written is recorded in error rather than screened, as a match alert failure already is.","endpoint":"/v1/batch","verification":{"steps":["B=$(curl -s -X POST https://api.noblesight.io/v1/batch -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"names\":[{\"full_name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\"},{\"full_name\":\"Nomenko\",\"entity_type\":\"entity\"}],\"lists\":[\"ofac\"]}' | jq -r .id); sleep 20; curl -s \"https://api.noblesight.io/v1/batch/status?id=$B\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '[.items[] | {name:.full_name, overall:.ownership.overall.status, alert:(.ownership_alert_id != null), reason:.alert_reason}]'"],"expect":"prints the Apex item with overall \"blocked\", alert true and reason \"ownership_control\", and the Nomenko item with overall \"review\", alert true and reason \"ownership_review\"; every screened item prints an overall status","audit":"each item's audit row (GET /v1/export?trace_id= with the batch's trace) stores the ownership object as returned, and its ownership alert (GET /v1/alerts/{id}) cites that row as its screening_id","regression":"TestBatchOwnership, TestBatchDecodesOwnership"}},{"date":"2026-11-07","category":"added","summary":"Persons OFAC blocks by programme definition (the blocked governments of Iran, North Korea, Syria, Venezuela, Cuba and the Crimea region) enter the us ownership determination: a state body a list's text names is reviewed as a blocked_government_candidate, a confirmed one blocks what it owns with blocked_basis on the hop and reviews what it controls","details":"OFAC's blocked-person universe is wider than its lists (FAQ 398): the Iran, North Korea, Syria, Venezuela, Cuba and Crimea programmes block persons who meet the definition of a blocked government without a designation, and the Iranian regulations reach any entity owned or controlled by the Government of Iran (31 CFR 560.304). Noble's ownership graph seeded its blocking frontier from list designations alone, so an unlisted state body blocked nothing and read clean. Now an unlisted owner a list's text names as a state body of one of those programmes is a candidate: its own us determination is review with reason blocked_government_candidate and the sentence that names it, and every entity it owns is review under the same reason, never clean. Noble's analysts confirm or reject a candidate; a confirmed person seeds the us regime at the next recompute, its own determination is blocked on its blocked_basis (government_of_iran, government_of_north_korea, government_of_syria, government_of_venezuela, government_of_cuba, crimea_region, each with its regulation), a stake of 50 percent or more it holds blocks the owned entity with blocked_basis and program on the hop, and a relationship with no stated percentage is review with reason controlled_by_blocked_government, because definition-based blocking reaches controlled entities and the percentage test does not. Every decision is recorded append-only with the analyst, the reason, the time and the regulation. No other regime changes. Candidates named by records already extracted appear at the ownership replay that follows the deploy, new ones at each import; confirmations take effect at the recompute that follows them.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Iranian State Retirement Fund\",\"entity_type\":\"entity\",\"lists\":[\"ofac\"],\"limit\":3}' | jq -c '.ownership.determinations.us | {status, reason, blocked_basis}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Maroun Oil \u0026 Gas Company\",\"entity_type\":\"entity\",\"lists\":[\"ofac\"],\"limit\":3}' | jq -c '.ownership.determinations.us | {status, reason, hops:[.chains[]?.hops[]?|{entity_name, is_sanctioned, blocked_basis, percent_owned}]}'"],"expect":"line 1 prints status \"review\" with reason \"blocked_government_candidate\" and blocked_basis \"government_of_iran\" (or status \"blocked\" with that blocked_basis once an analyst has confirmed the fund), never \"clean\" or \"unknown\"; line 2 prints status \"review\" and a hop whose entity_name contains \"National Iranian South Oil Company\" with blocked_basis \"government_of_iran\" (reason \"blocked_government_candidate\" before confirmation, \"controlled_by_blocked_government\" after), never \"unknown\"","audit":"every candidate and decision is held append-only in ownership.program_blocked with its basis, regulation, evidence assertion, analyst, reason and time, readable by the operator at /v1/admin/ownership/program-blocked; the determination and its hops carry the basis in ownership.derivative_block and the screening record (GET /v1/export?trace_id=)","regression":"TestBlockedGovernmentBasis, TestBlockedGovernmentCandidateFromText, TestProgramBlockedRows, TestSettleSticky"}},{"date":"2026-11-07","category":"added","summary":"A review-grade ownership determination raises its own alert: alert_reason ownership_review, change_type ownership_review, score 60, at screen time and from portfolio monitoring when an enrolled entity newly comes under review","details":"A review determination is correct under OFAC when a list discloses a relationship to a blocked person but no stake (FAQ 398: control without 50 percent ownership is not blocking, but may be a red flag), and it used to be inert: a field in the response, no alert, no monitoring event, no SLA. It now raises an ownership alert of its own, distinct from the blocking alert: alert_reason ownership_review, change_type ownership_review, score 60 (below the four-eyes line, on the 72-hour SLA), with the determination and its chains as evidence, on POST /v1/screen when overall.status is review, and from portfolio monitoring when the nightly recompute newly places an entity a monitored name resolves to under review. One open ownership alert per determined entity and change type, so a review alert and a later blocking alert on the same entity are two cases, and a block that later falls to review keeps its open blocking alert and opens no second case. The blocking alert is unchanged. The webhook event's data now carries alert_reason and change_type beside score and source, so a consumer routes or counts review alerts apart without fetching the alert.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Nomenko\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{status:.ownership.overall.status, ownership_alert_id, alert_reason}'","ID=$(curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Nomenko\",\"entity_type\":\"entity\",\"limit\":3}' | jq -r .ownership_alert_id); curl -s https://api.noblesight.io/v1/alerts/$ID -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{source, score, four_eyes_required, alert_reason, status:.ownership.overall.status, chains:(.ownership.determinations|[.[].chains[]?]|length)}'"],"expect":"line 1 prints status \"review\", a non-null ownership_alert_id and alert_reason \"ownership_review\"; line 2 prints source \"ownership\", score 60, four_eyes_required false, alert_reason \"ownership_review\", status \"review\" and chains at least 1; a second run of line 1 prints the same ownership_alert_id","audit":"the alert row carries change_type ownership_review, its evidence (GET /v1/alerts/{id}) the determination as returned, and the screening record (GET /v1/export?trace_id=) the same; a monitoring-raised review alert carries portfolio_entity_id and a trigger of kind ownership_change naming the recompute run","regression":"TestScreenOwnershipEndToEnd, TestOwnershipRecomputeAlertsPortfolio"}},{"date":"2026-11-07","category":"changed","summary":"An analyst's rejection of an extracted ownership assertion removes it from every determination at the next recompute and is never resurrected by a later extraction; a percentage stated in a list's own text now carries confidence high","details":"Noble's ownership graph reads ownership statements from the sanctions lists' published text, quoted verbatim on each hop. Until now the review queue those assertions sat in could not be cleared and a rejection could not take effect. Noble's analysts now record a verdict on an assertion with the reason and the time: a rejected assertion contributes to no determination from the next recompute on, its edge is retired from the graph at the next import, and a later extraction of the same record's same relationship is recorded rejected again and never edged, so a corrected reading stays corrected across publishes. A stake the list's own text states is the authority's statement and counts under the regime's rule without prior approval, as it always did; its hop now carries confidence \"high\", the grade the contract defines for an explicit percentage from an authoritative source, instead of \"low\". A relationship stated without a percentage keeps \"low\" and remains review. Hops change at the ownership replay that follows the deploy.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.determinations[]?.chains[]?.hops[]|select(.source|startswith(\"list_text:\"))|{percent_owned, confidence}]'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Nomenko\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.determinations[]?.chains[]?.hops[]|select(.source|startswith(\"list_text:\"))|{percent_owned, confidence}]'"],"expect":"line 1 prints hops with a percent_owned and confidence \"high\"; line 2 prints hops with no percent_owned and confidence \"low\"; and a determination never carries a hop whose assertion Noble's analysts have rejected (the rejected assertion's sentence no longer appears on any screening of the owned name after the next nightly recompute)","audit":"every verdict is held append-only in ownership.assertion_verdict with the analyst, reason and time and is readable by the operator at /v1/admin/ownership/assertions/{id}; the recompute that dropped the determination is logged in ownership.recompute_log and ownership.derivative_block_changelog","regression":"TestAdminOwnershipPending, TestOwnershipVerdictInheritance"}},{"date":"2026-11-07","category":"added","summary":"Every ownership hop extracted from a list's text cites the statement it rests on: hops[].assertion_id and hops[].source_sentence carry the audit-store row and the verbatim sentence the stake came from","details":"A hop whose source is list_text:\u003clist\u003e now carries assertion_id, the identifier of the extracted ownership statement in Noble's audit store (the list record, its publish id, the verbatim quote and its review status), and source_sentence, the sentence of the list's published text the percentage was read from. An examiner reads the stake against the sentence on the determination itself and on the alert that carries it, instead of joining list, keys and dates by hand; the operator resolves the id to the full row at GET /v1/admin/ownership/assertions/{id}. Hops from GLEIF corporate-relationship data and OFAC Linked To pointers carry neither. Determinations gain the fields at the ownership replay that follows the deploy, which restates every list's persisted assertions onto the graph with their ids, and the recompute after it.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.determinations[]?.chains[]?.hops[]|select(.source|startswith(\"list_text:\"))|{assertion_id:(.assertion_id != null), sentence:(.source_sentence|length \u003e 0), percent_owned}]'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"PJSC AFK Sistema\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.determinations[]?.chains[]?.hops[]|select(.source|startswith(\"list_text:\"))|[.assertion_id != null, (.source_sentence|length \u003e 0)]]|flatten|unique'"],"expect":"line 1 prints at least one hop and every hop prints assertion_id true, sentence true and a percent_owned; line 2 prints exactly [true]","audit":"the screening record (GET /v1/export?trace_id=) stores the hops with their assertion ids and sentences as returned; an ownership alert's evidence (GET /v1/alerts/{id}) carries the same; the cited row, its publish id and review status are held in ownership.extracted_relation, readable by the operator at /v1/admin/ownership/assertions/{id}","regression":"TestAssembleChainsIndirect, TestAttachSentences, TestAdminOwnershipPending, TestScreenDecodesOwnership"}},{"date":"2026-11-07","category":"changed","summary":"A screening never fails open on its alert or its ownership control: a result that must carry an alert and cannot is refused with 503 alert_unavailable and nothing of it is committed, and a screening whose ownership determination could not be read says so (ownership.overall.status \"unavailable\", an ownership_unavailable warning and coverage issue)","details":"Two edges used to pass silently. When the ownership determination could not be read, the response carried no ownership object at all, indistinguishable from a server without the control, and the stored record read as a screening the 50 Percent Rule question was never asked of. Now the object is always present on a live name screening: ownership.overall.status is \"unavailable\", every selected regime's determination says the same under its rule, warnings carries ownership_unavailable, and coverage carries an issue with list \"ownership\" and code ownership_unavailable with degraded true and no election. The name matches are complete; the ownership question is unanswered and the screening should be repeated. When a screening produces a result that must carry an alert, a name match at or above your threshold or a near miss, or a blocked ownership determination, and the alert cannot be written, the screening now fails with 503 alert_unavailable, with request_id and trace_id. The screening record, the alert, its webhook and its audit row commit together or not at all, so no completed record and no alert exists for a screening you were told failed, and no hit exists that nobody was told about; the attempt is recorded as a failed screening. Retry the screening.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{present:(.ownership != null), overall:.ownership.overall.status, ownership_alert_id, degraded:.coverage.degraded, issues:[.coverage.issues[].code]}'","for i in 1 2 3 4 5 6 7 8 9 10; do curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d \"{\\\"name\\\":\\\"Example Holding $i\\\",\\\"limit\\\":1}\" | jq -c '[(.ownership != null), (.ownership.overall.status), ([.warnings[]?.code]|index(\"ownership_unavailable\") != null)]'; done"],"expect":"line 1 prints present true, overall \"blocked\", a non-null ownership_alert_id, degraded false and an issues list without \"ownership_unavailable\"; every line of step 2 prints an ownership object present with an overall status, and wherever that status is \"unavailable\" the same line prints true for the ownership_unavailable warning (a live screening never has a null ownership object with no warning naming it)","audit":"the screening record (GET /v1/export?trace_id=) stores the ownership object with status unavailable, the warning and the degraded coverage exactly as returned; a screening failed with alert_unavailable is stored only as a failed attempt under its request_id and trace_id with the error as its response, with no completed record and no alert","regression":"TestScreenOwnershipResolverFaultIsNamed, TestScreenBlockedWithoutAlertFailsClosed"}},{"date":"2026-11-07","category":"changed","summary":"The ownership determination is made per sanctioning authority under that authority's own rule: ownership.determinations.\u003cregime\u003e (us, uk, eu, un, ca, au) with a derived ownership.overall, schema 2; lists selects the regimes and the alert follows the selection","details":"Every authority has a derivative-ownership rule and they differ: OFAC blocks at 50 percent or more in the aggregate across its programmes and a block is sticky once reached (FAQs 398, 401, 402); OFSI requires more than 50 percent and does not sum different designated persons' holdings; the EU and Canada attach at 50 percent or more, aggregated; the UN and Australia test ownership or control without a numeric threshold. The ownership object now carries one determination per regime, each reached under that regime's rule over the owners that regime designates and naming the rule it was reached under (rule), the regime's consequence word as its status (blocked under us, owned_or_controlled under uk, eu and un, deemed_owned under ca, controlled_asset under au), the figure the rule tested (aggregate_percent), whether a link without a disclosed stake is involved (control), the date the block began (blocked_since), a reason when the threshold alone did not decide it (ownership_reduced_after_block: a stake that fell after a block stays blocked under us and becomes review elsewhere, never clean without a human verdict), and that regime's evidence chains. An entity designated under one regime is still evaluated under the others, so an EU-listed company wholly owned by an OFAC-designated person reads designated under eu and blocked under us. overall is derived over the regimes the request's lists selected and carries no rule of its own; the ownership alert is raised exactly when overall.status is blocked, so a request scoped to ofac is never alerted on a block only a foreign authority's rule reaches, and a request without lists reads every regime. The top-level status, cumulative_percent, blocked_by_lists and chains are removed; schema is 2. Evidence on alerts raised before this change is rendered under the legacy regime with rule pre_regime_union, exactly as it stood. Determinations change at the first recompute after the deploy; that recompute rewrites the determination table per regime and raises no portfolio-monitoring event for the rewrite itself.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"PJSC AFK Sistema\",\"entity_type\":\"entity\",\"lists\":[\"ofac\"],\"limit\":3}' | jq -c '{schema:.ownership.schema, ownership_alert_id, overall:.ownership.overall, regimes:(.ownership.determinations|keys), us:.ownership.determinations.us.status}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"PJSC AFK Sistema\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{overall:.ownership.overall, eu:(.ownership.determinations.eu|{rule,status,aggregate_percent}), uk:.ownership.determinations.uk.status}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{overall:.ownership.overall, us:(.ownership.determinations.us|{rule,status,aggregate_percent,blocked_since}), owner:[.ownership.determinations.us.chains[0].hops[0].designations[].list]}'"],"expect":"line 1 prints schema 2, ownership_alert_id null, a determinations key list of exactly [\"us\"], and us not \"blocked\" (Sistema's co-owners are designated by France and the UK, not OFAC); line 2 prints an eu determination with rule \"eu_ownership_or_control\", status \"owned_or_controlled\" and aggregate_percent 64.4 (the EU rule sums the two holdings), uk \"clean\" (OFSI does not sum them and neither exceeds 50), and overall.regimes containing \"eu\"; line 3 prints overall.status \"blocked\" with regimes [\"us\"], a us determination with rule \"ofac_sdn_50_percent\", status \"blocked\", aggregate_percent 90 and a blocked_since date, and an owner designation list containing \"ofac\"","audit":"the screening record (GET /v1/export?trace_id=) stores the ownership object as returned, every regime's determination and chains included; the per-regime determination and every change to it are held in ownership.derivative_block and the append-only ownership.derivative_block_changelog with the run that made it in ownership.recompute_log, each row naming its regime and rule; an alert's evidence (GET /v1/alerts/{id}) is the object as it stood when the alert was raised","regression":"TestEvaluateRegimeTable, TestEvaluateContributions, TestEvaluateReviewAndPartial, TestSettleSticky, TestLegacyRegime, TestDiffDeterminations, TestOwnershipResolver, TestScreenOwnershipEndToEnd, TestScreenDecodesOwnership, TestGetAlertDecodesOwnershipEvidence"}},{"date":"2026-11-06","category":"changed","summary":"A designated party is one entity in the ownership graph whatever lists designate it: ownership hops and the screened entity carry designations[] and identity_grade, an owner's entity_key is ent:\u003cid\u003e rather than one list's record, and a stake it holds is counted once","details":"The ownership graph kept one node per list record, so a person OFAC, the EU, France and the UK all designate was four owners: a stake a list's narrative stated reached whichever record the extraction landed on, blocked_by_lists named that one list, and the same party could in principle be counted once per list toward the aggregate the rule sets. Designations are now resolved to one entity per party before the graph is built, and merged only on graded evidence: a shared published reference number (reference); a name equal under the engine's fold corroborated by an equal date of birth, a shared identification number or, for a company, two distinct name forms in common (strong); or an analyst's recorded verdict (verdict). A name alone, equal or fuzzy, never merges: such pairs wait in the identity review queue, so a parent and a subsidiary that share an alias stay two entities. Every hop whose owner is designated now carries designations[] (each list's record, with the grade by which it joined) and identity_grade (the entity's weakest link; singleton for one designation); sanctioned_list still lists the same lists. The screened name's own entity is reported as ownership.entity with its designation set, on a determination and also on an unknown determination when the name is exactly a designated party's (such a party is not evaluated by the computation; its designation is the finding). The entity_key of a designated owner changes value from a list record (sdn:29046, france:2112) to its entity (ent:\u003cid\u003e); keys are opaque identifiers and no field changes shape, so no version is minted. Ownership alerts are deduplicated on entity_key, so the first screening of an already alerted entity after this change may raise one further alert under its new key. Delisting under one authority removes that designation and leaves the entity. Persisted ownership assertions move onto the entities at the ownership replay that follows the deploy, and determinations change at the recompute after it; the ownership import now refuses to run when identity resolution fails rather than seed a graph keyed two ways.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Apex Development and Projects LLC\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '[.ownership.chains[].hops[0]]|unique|map({owner:.entity_key, designations:([.designations[]|\"\\(.list):\\(.uid)\"]|sort), grade:.identity_grade})'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Naftiran Intertrade Company SRL\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{status:.ownership.status, entity:([.ownership.entity.designations[]|\"\\(.list):\\(.uid)\"]|sort)}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Hassan Jamaleddine\",\"entity_type\":\"individual\",\"limit\":3}' | jq -c '{top:.summary.top_match.uid, designations:[.ownership.entity.designations[]|\"\\(.list):\\(.uid)\"]}'"],"expect":"line 1 prints exactly one owner whose entity_key starts with \"ent:\", whose designations include \"ofac:29046\", \"france:2112\", \"eu:119304\" and \"uk:SYR0258\", and whose grade is \"reference\" or \"strong\" (never \"candidate\"); line 2 prints an entity whose designations include \"ofac:11319\" and not \"ofac:11318\" (the parent), whatever the status; line 3 prints top \"20561\" and designations that contain neither \"ofac:29046\" nor \"france:2112\"","audit":"the screening record (GET /v1/export?trace_id=) stores the ownership determination as returned, chains and designations included; the membership behind every designation, the grade and evidence of each merge and the review queue are held in ownership.identity_member, ownership.identity_log and ownership.identity_candidate, readable by the operator at /v1/admin/ownership/identity/\u003ckey\u003e; the recompute that changed a determination is logged in ownership.recompute_log and ownership.derivative_block_changelog","regression":"TestOwnershipIdentity, TestAssembleChainsDesignations, TestRekeyPrior, TestIdentityGrade, TestEntityKeyRoundTrip"}},{"date":"2026-11-05","category":"fixed","summary":"The 50 Percent Rule ownership graph resolves an owner by every name its list publishes: aliases are indexed, names are compared under the engine's fold, and a trailing acronym in brackets is ignored, so a stake a list states under an a.k.a. lands on the designated owner and can block","details":"An ownership statement in a list's narrative (\"X holds a majority of Y\") names the owner as the list writes it, which is often an alias, a spelling with different punctuation, or a name followed by its acronym in brackets. The graph resolved that name to a designation only when it equalled the designation's primary name exactly, lowercase. Any other spelling minted a free-floating node that no list designates, and a chain through it produced nothing: not blocked, not review. On production that stranded over a thousand percentage-bearing assertions, among them dozens of owners that are active designations' aliases, and no blocked determination rested on an OFAC-designated owner. Three changes: every alias a list publishes for a designation is indexed beside its primary name, except aliases the list itself grades weak; names are compared under the fold the matching engine applies to screened names (case, diacritics, punctuation, width, spacing); and an owner written as \"Name (ACRONYM)\" is also tried as \"Name\", where the bracket holds an abbreviation and not a qualifier such as a country. Where a designation and a stranded node share a name, the designation now wins the screen-time tie, so a designated party is no longer reported ownership clean through its twin (\"Rosneft\", SDN 17022, was). Index rows an earlier normalisation produced leave on the next sync. Existing assertions are repaired by the ownership replay job, which re-resolves every persisted owner and owned name against the richer index without re-extraction, retires the edge an assertion held under its old nodes, and logs each move; determinations change at the recompute that follows it. A determination that was blocked stays blocked; new blocks appear where an aliased owner's stake now reaches its designation. No field changes shape; cross-list identity (which list's node an owner lands on when several designate it) is #705.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Rosneft\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{top:.summary.top_match.uid, status:.ownership.status, entity_key:.ownership.entity_key}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"GSK\",\"entity_type\":\"entity\",\"limit\":3}' | jq -c '{status:.ownership.status, pct:.ownership.cumulative_percent, owner:(.ownership.chains[0].hops[0]|{entity_key,percent_owned,is_sanctioned})}'"],"expect":"line 1 prints top 17022 with a status other than clean and no entity_key xname:rosneft (a designated party is unknown under the ownership control, its name match carries the alert); line 2 prints status blocked, pct 60 and an owner hop whose entity_key is a list designation (not xname:), percent_owned 60, is_sanctioned true, after the ownership replay and recompute that follow the deploy","audit":"the screening record (GET /v1/export?trace_id=) stores the ownership determination as returned, with its chains, as_of and source_publish_ids; the re-keyed assertions keep their original extraction timestamp and verbatim quote in ownership.extracted_relation, and the recompute that changed a determination is logged in ownership.recompute_log and ownership.derivative_block_changelog","regression":"TestOwnershipResolver, TestOwnershipRetrievalSuperset, TestOwnershipNeighborhoodDecidesAsResolve, TestNormalizeName, TestStripTrailingParenthetical"}},{"date":"2026-11-04","category":"fixed","summary":"GET /v1/portfolio returns date_of_birth as YYYY-MM-DD, the form the contract states and the form it accepts, not a midnight timestamp","details":"A monitored entity's date_of_birth is a date: POST /v1/portfolio and POST /v1/screen accept it as YYYY-MM-DD and the contract documents it so. The portfolio reads rendered it as a midnight timestamp (1970-01-01T00:00:00Z), so a caller that stored 1970-01-01 read back a different string, and a comparison against the value it sent failed. Every portfolio read now returns the date as stored, YYYY-MM-DD, on GET /v1/portfolio, on the entity POST and PUT /v1/portfolio answer with, and on GET /v1/alerts/{id} entity. Confidential Portfolio entities were unaffected: their decrypted value was already the date as sent. Found closing #623 against production. No field changes meaning and no version is minted.","endpoint":"/v1/portfolio","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/portfolio -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"Dob Format Probe 2026\",\"date_of_birth\":\"1970-01-01\",\"country\":\"MX\"}' | jq -c '{date_of_birth, country}'","curl -s \"https://api.noblesight.io/v1/portfolio?q=Dob%20Format%20Probe%202026\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.data[0] | {date_of_birth, country}'"],"expect":"both lines print {\"date_of_birth\":\"1970-01-01\",\"country\":\"MX\"}; remove the probe with DELETE /v1/portfolio?id= afterwards","audit":"the enrollment's screening record (GET /v1/export?trace_id=) carries the request's date_of_birth as sent; the stored column is a date, so no value changed, only its rendering","regression":"TestScreenEnrollmentReported, TestPortfolioWriteScreens"}},{"date":"2026-11-04","category":"fixed","summary":"A match's pathway detail and enrichment render the same bytes wherever the match appears: a name with more than one screened form keeps every form's match_sources detail when the response is cut at limit, and programs, addresses, ids and nationalities are listed in the order the publisher lists them","details":"Noble proves that a portfolio monitoring hit is the POST /v1/screen hit for the same customer and listed party. Routing that proof through the live handler found two places where the same pair rendered differently. First, a name screened in more than one form (one with a bracketed note, one carrying invisible characters, one with mojibake repaired) was cut to the response's limit per form before the forms were merged, so on a crowded name the returned match could carry a pathway's trigram_score from one form while the pair decision, which cuts nothing, reported the other's; the record's score and its presence in the response were never affected. The cut now runs on the merged ranking, so a returned match carries every form's detail. Second, the programs, addresses, ids, nationalities and citizenships of an OFAC match, and the addresses, identifiers and nationalities of a UK, EU, UN, France or Belgium match, were read without an order and could come back in a different order from one screening to the next, or between a screening and a monitoring alert for the same party. Each is now listed in the order the publisher lists it. No score, verdict, alert or field changes meaning; a response compares equal to a monitoring alert's match_payload under jq -S.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"lists\":[\"ofac\"]}' | jq -c '.matches[]|select(.uid==\"35096\")|{programs,ids:[.ids[].number]}' \u003e a.json","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin (c/o Kremlin)\",\"lists\":[\"ofac\"],\"limit\":1}' | jq -c '.matches[]|select(.uid==\"35096\")|{programs,ids:[.ids[].number],sources:[.match_sources[]|{source,trigram_score}]}' \u003e b.json","for i in 1 2 3; do curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"lists\":[\"ofac\"]}' | jq -c '.matches[]|select(.uid==\"35096\")|{programs,ids:[.ids[].number]}'; done | sort -u | wc -l"],"expect":"a.json and b.json list the same programs and ids in the same order; b.json's sources carry the stripped form's scores (ofac_aka at 100, the alias Vladimir Putin, and ofac_sdn at 71.4) although the request's limit was 1 and the name carried a bracketed note, which on its own scores far lower; the third line prints 1, three identical renderings","audit":"GET /v1/export?trace_id= returns each screening with the matches as rendered; a monitoring alert's match_payload (GET /v1/alerts/{id}) for the same party compares equal to the screening's match under jq -S, less the per-match trigger key","regression":"TestReverseHitEqualsScreen, TestMonitoringEndToEnd, TestScreenRecordEqualsScreen"}},{"date":"2026-11-04","category":"added","summary":"A monitoring alert raised by a changed designation says what changed: trigger.changes lists what the publish added, removed or changed in it, read from the list's own before-and-after record, and caused_hit marks the change the customer's match came through","details":"A monitoring alert said why it opened (a list change, a re-screen) but not what the list changed, so an analyst explaining an alert on a customer screened clean months ago had to join the alert to the publish and the list's changelog by hand. A list_change alert on a modified designation now carries trigger.changes: each alias, name, native-script name, program, type, title, remark, date of birth or other field the publish added, removed or changed, as {field, op, value, previous}, read from the list's own record of the designation before and after that publish. caused_hit is true on the change the alert's match came through: the match's matched_text is a name that change added, whole or one part of it such as the alias's surname, never a name the designation already listed. An alert raised because a new alias matched reads differently from one where a name already listed matched and only a program changed. An alias whose strong/weak category or other attribute changed is a changed alias, with its attributes in brackets in value and previous. changes is absent on an added designation (all of it is new), on re-screen and re-probe alerts, and where the list keeps no before and after for the publish, never an empty list. An OFAC native-script name added by its Advanced feed is a native_name change. The screening record behind the alert carries the trigger; when one customer matched the same listed record more than once, each match in the record carries its own alert's trigger. Additive, so no version is minted.","endpoint":"/v1/alerts/{id}","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/alerts?source=monitoring\u0026limit=100\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.data[]|select(.trigger.change_type==\"modified\")|.id' | head -1","curl -s https://api.noblesight.io/v1/alerts/$ID -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{changes: .trigger.changes, matched: [.match_payload[].match_sources[].matched_text]}'"],"expect":"for a monitoring alert created after 2026-11-04 on a modified OFAC designation, changes has at least one entry, each with field, op and value; every entry with caused_hit true added a name equal (word for word, case aside) to one of matched or to one part of one, and no name already listed before the publish is credited; an added designation's alert has no changes key","audit":"GET /v1/export?trace_id= with the alert's trace_id returns the monitoring record, whose response.trigger carries the same changes; the list's changelog for the trigger's publish_id holds the before and after they were read from","regression":"TestMonitoringAlertTriggerChanges, TestMonitoringAlertTriggerRegistryList, TestMonitoringAlertTriggerNativeName, TestDiffSnapshots, TestWithCausedHit, TestGetAlertDecodesTriggerChanges"}},{"date":"2026-11-03","category":"added","summary":"Portfolio monitoring's full re-screen covers ownership: every monitored portfolio entity is re-checked against every entity currently blocked under the 50 Percent Rule, not only those a recompute newly blocks, and alerts any that were never alerted","details":"Ownership monitoring reacted only to an entity a recompute newly blocked. An entity blocked before ownership monitoring shipped, or a customer name that newly resolves to an already-blocked entity through another of its names, raised nothing until that customer was screened again. Noble's full portfolio re-screen, which it runs after changes to its matching or monitoring rules, now covers every blocked ownership determination as of the latest recompute, decided exactly as at screen time. It never raises a pair that already has an ownership alert for the same blocked entity, open or closed, including one raised when the same name was screened; that alert gets a list_event activity entry saying the re-screen reached it. An earlier alert about a different entity never suppresses one. Its alerts are source ownership with a trigger of kind engine_change, list ownership and publish_id the recompute run. Additive, so no version is minted.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/alerts?source=ownership\u0026limit=100\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.data[]|select(.trigger.kind==\"engine_change\" and .trigger.list==\"ownership\")|{id, portfolio_entity_id, screened_name, publish_id: .trigger.publish_id}'","curl -s https://api.noblesight.io/v1/alerts/$ID -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{status: .ownership.status, entity_key: .ownership.entity_key, chains: (.ownership.chains|length), screening_id}'"],"expect":"every alert step 1 lists has a portfolio_entity_id; step 2 for it returns status blocked, a non-empty entity_key, at least one chain and a positive screening_id; POST /v1/screen of its screened_name returns ownership.status blocked with the same entity_key","audit":"each alert names its monitoring evidence row in screening_id, and its trigger names the recompute run its determination was taken from; a pair the re-screen skipped carries a list_event on its earlier alert","regression":"TestOwnershipFullRescreen"}},{"date":"2026-11-03","category":"added","summary":"Portfolio monitoring covers the 50 Percent Rule: when Noble's nightly ownership recompute newly determines an entity blocked, every monitored portfolio entity whose name POST /v1/screen would block for it raises an ownership alert, with the determination as evidence and a trigger of kind ownership_change","details":"Monitoring reacted only to list publishes. An entity that came to be majority-owned by sanctioned parties overnight was blocked at screen time but raised nothing on a portfolio already enrolled, so the customer learned of it only by re-screening that name. Each recompute now records every determination it changes (newly blocked, newly review, moved between them, dropped) in an append-only record written with the determinations themselves, and each newly blocked entity is screened against every monitored portfolio, Confidential Portfolio entities included, within minutes of the recompute. The decision is the one POST /v1/screen makes: an entity is alerted when screening its name returns ownership status blocked for the newly blocked entity. The alert has source ownership, alert_reason ownership_control, score 100 and the portfolio_entity_id; ownership carries the determination with its chains; trigger has kind ownership_change, list ownership, publish_id the recompute run and imported_at when the run recorded the change; it goes out through the same webhook outbox as every alert. Ownership alerts now deduplicate on the determined entity (ownership.entity_key, a new field on the ownership object) for as long as the alert is open, per portfolio entity for monitoring, instead of on the screened name within five minutes: the same blocked entity screened again a week later, or under another of its names, returns the open alert, and a recompute that blocks it again while open is recorded as a list_event on it. Newly review determinations are recorded but never alert, as at screen time. A decrypted Confidential Portfolio name is decided in memory, never sent to the database. Not covered yet: entities already blocked before this change, and a name that newly resolves to an already-blocked entity; a one-time sweep of every blocked entity follows. Additive, so no version is minted.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/alerts?source=ownership\u0026limit=100\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.data[]|select(.trigger.kind==\"ownership_change\")|{id, screened_name, alert_reason, portfolio_entity_id, trigger: (.trigger|{kind, list, publish_id})}'","curl -s https://api.noblesight.io/v1/alerts/$ID -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{source, alert_reason, status: .ownership.status, entity_key: .ownership.entity_key, pct: .ownership.cumulative_percent, owner: .ownership.chains[0].hops[0].entity_name, trigger: .trigger.kind}'"],"expect":"after a recompute that newly blocks an entity whose name a monitored portfolio entity carries, step 1 lists an alert with alert_reason ownership_control, a portfolio_entity_id and trigger kind ownership_change on list ownership; step 2 with that id returns source ownership, ownership.status blocked, a non-empty entity_key, cumulative_percent \u003e= 50 and a named owner on the first chain","audit":"the trigger's publish_id is the recompute run; that run's row in Noble's append-only ownership changelog holds the determination as it was recorded, and the alert's ownership object is frozen at creation, so both survive later recomputes; the alert's screening_id is the monitoring decision's record, which GET /v1/export?trace_id= with its trace_id returns","regression":"TestOwnershipRecomputeAlertsPortfolio, TestOwnershipRecomputeAlertsEncryptedPortfolio, TestOwnershipNeighborhoodDecidesAsResolve, TestOwnershipRetrievalSuperset, TestDiffDeterminations, TestAuditImmutability"}},{"date":"2026-11-03","category":"fixed","summary":"An ownership alert carries its evidence: GET /v1/alerts/{id} returns the ownership determination it was raised on, chains included, and every alert returns the screening_id and trace_id of the screening that raised it","details":"An ownership alert is categorical, four-eyes and on a 24-hour SLA, and it arrived with match_count 0, an empty match_payload and nothing else: no sanctioned owner, no list, no percentage, and no pointer back to the screening, so the only route to the evidence was to screen the name again against today's data. The determination was stored with the alert and never returned. GET /v1/alerts/{id} (and every alert mutation's response) now returns ownership on ownership alerts: status, entity_name, entity_key, cumulative_percent, blocked_by_lists, every chain with its hops, as_of and source_publish_ids, exactly as when the alert was raised, including on alerts raised before today. Every alert, in the list and the detail view, also returns trace_id and, except batch alerts, screening_id: the screening_id POST /v1/screen returned, or the monitoring decision's record (GET /v1/export?trace_id= returns it). The MCP get_alert tool returns the same fields. Additive, so no version is minted.","endpoint":"/v1/alerts/{id}","verification":{"steps":["ALERT=$(curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Advanced Mining Works\"}' | jq -r '.ownership_alert_id')","curl -s https://api.noblesight.io/v1/alerts/$ALERT -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{status: .ownership.status, pct: .ownership.cumulative_percent, lists: .ownership.blocked_by_lists, owner: .ownership.chains[0].hops[0].entity_name, owner_list: .ownership.chains[0].hops[0].sanctioned_list, screening_id, trace_id}'","curl -s \"https://api.noblesight.io/v1/export?trace_id=$TRACE_ID\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{id, trace_id}'"],"expect":"step 2 prints status blocked, pct \u003e= 50, a non-empty lists array, a non-empty owner and owner_list, a positive screening_id and a non-empty trace_id; step 3, with TRACE_ID set to that trace_id, prints an id equal to that screening_id","audit":"the ownership object is the alert's stored match_details, which the database freezes when the alert is created; screening_id is the alert's screening record, exported by its trace_id","regression":"TestOwnershipAlertEvidence, TestGetAlertDecodesOwnershipEvidence, TestScreenOwnershipEndToEnd"}},{"date":"2026-11-03","category":"fixed","summary":"The monitoring_latency control counts list changes still waiting for their portfolio re-screen: latency.unfinished says how many there are, and one past the 900-second budget is measured at its age so far and keeps the control from reading active","details":"The monitoring_latency control on GET /v1/account/coverage measured only list changes whose portfolio re-screen had finished. A change still waiting to be queued, or with part of its re-screen pending, retrying, failed or parked for operator review, had no finish to measure and was left out, so a change outstanding for days could not appear until its re-screen completed, and the control could read active through exactly the delay it exists to report. Every change in the window now counts. latency.unfinished is the number not yet fully re-screened (0 in a healthy state). One younger than the budget is in flight and not yet measured. One older than the budget is late whatever happens next: it enters the distribution and changes at its age so far, fails the budget, overrides any partial webhook delivery on the delivered basis, and keeps the control from reading active, with guidance naming how many are outstanding. A change resolved with nothing to screen is finished, not unfinished. latency.unfinished is additive, so no version is minted.","endpoint":"/v1/account/coverage","verification":{"steps":["curl -s https://api.noblesight.io/v1/account/coverage -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.controls[]|select(.id==\"monitoring_latency\")|{status, latency: (.latency|{changes, unfinished, p95_s, max_s})}'","curl -s https://api.noblesight.io/v1/lists -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.lists[]|select(.last_change != null)|{list, detected: .last_change.detected_at, rescreened_after_s: .last_change.rescreened_after_s}'"],"expect":"step 1 prints a numeric latency.unfinished; whenever step 2 shows a list whose detected instant is more than 900 seconds old with rescreened_after_s null, step 1 prints unfinished of at least 1 and a status other than active","audit":"an unfinished change is a delta outbox row the drain has not resolved, or one whose own reverse-screen job rows (same trace id) are not all completed; both tables keep their rows under the standard retention rules, so the count can be reconciled change by change","regression":"TestLatencyCountsUnfinishedChanges, TestLatencyCountsOnlyTheChangesOwnWork, TestCoverageMonitoringLatencyControl"}},{"date":"2026-11-03","category":"fixed","summary":"Monitoring latency measures each list change on its own work: a later engine re-screen, variation backfill or operator reprobe no longer stretches a change's re-screen hop on GET /v1/lists or the monitoring_latency control","details":"The re-screen hop on GET /v1/lists (last_change.job_finished_at, rescreened_after_s) and the monitoring_latency control on GET /v1/account/coverage attributed every reverse-screen job against a list version to the change that produced that version. Engine re-screens, AI name-variation backfills and operator reprobes run against the list's current version too, so each one that ran after a change moved that change's re-screened instant to its own finish: a UK change re-screened in 23 seconds read as 188,192, and the 90-day control's p95 read 369,008 seconds while every list's latest change had been re-screened in under a minute. Both surfaces now measure a change on the jobs queued for it and the alerts those jobs raised, and nothing else. A job of the change's own that stalled and was retried or requeued still counts at its real finish, so the fix can only remove work that was never the change's, never hide a slow change: the October OFAC delta that sat parked for 19 days until the 2026-11-01 requeue still counts at its full 19 days. The delivered basis is tightened the same way: two changes on one list version (an OFAC SDN delta and its native-name growth) no longer share each other's first delivery. Values change, shapes do not, so no version is minted.","endpoint":"/v1/account/coverage","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/account/coverage?from=2026-10-30\u0026to=2026-11-02\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.controls[]|select(.id==\"monitoring_latency\")|.latency|{changes, p50_s, p95_s, max_s, basis}'","curl -s https://api.noblesight.io/v1/lists -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.lists[]|select(.last_change != null)|{list, detected: .last_change.detected_at, finished: .last_change.job_finished_at, rescreened_after_s: .last_change.rescreened_after_s}'"],"expect":"step 1 prints changes 2, basis rescreened and p95_s under 60 (the France and UK changes of 2026-10-30, re-screened in about 29 and 23 seconds); step 2 prints, for every list whose last change predates this entry, a finished instant within a minute of detected and rescreened_after_s under 60","audit":"each change's jobs are the reverse-screen job rows the delta drain queued under the change's own trace id, all chunks of one change sharing it; a requeued job keeps its row and trace, so its late finish stays on the change it belongs to","regression":"TestLatencyCountsOnlyTheChangesOwnWork, TestListsLastChange, TestCoverageMonitoringLatencyControl"}},{"date":"2026-11-02","category":"added","summary":"The monitoring chain proves itself every hour: a Noble-published canary feed flips hourly, and the monitoring_latency control passes only while the latest flip was delivered within the 900-second budget AND every real list is fresh","details":"Real list changes land most days on some lists and rarely on others, so a quiet day left the notify-within-minutes promise unproven and a dead chain invisible until the next real change. Noble now publishes its own tenth source — a one-row feed flipped every hour — enrolled in a Noble-held tenant whose webhook receiver measures how long each flip took to arrive, exercising every hop a real change rides: probe, download, import, delta queue, portfolio re-screen, alert, webhook delivery. The canary is internal and invisible as a list: it cannot be named on /v1/screen (indistinguishable from a list that does not exist), never appears in lists_screened, list_versions, coverage, exports or GET /v1/lists, and is reverse-screened only against the Noble-held tenant. Its measurement is public: the monitoring_latency control on GET /v1/account/coverage gains latency.canary — the latest flip, its delivered_after_s, and within_budget — plus latency.lists_fresh, and while the canary publishes, the control passes only when the canary is in budget AND every customer-visible list is fresh, so a green distribution can never mask a dead chain and a green canary can never mask a red publisher. Synthetic flips are excluded from the latency distribution itself. Additive, so no version is minted.","endpoint":"/v1/account/coverage","verification":{"steps":["curl -s https://api.noblesight.io/v1/account/coverage -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.controls[]|select(.id==\"monitoring_latency\")|{status, canary: .latency.canary, lists_fresh: .latency.lists_fresh}'","curl -s https://api.noblesight.io/v1/lists -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '[.lists[].list]'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"ZZCANARY NOBLE\",\"lists\":[\"canary\"]}' | jq -c '{type, code, param}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"ZZCANARY NOBLE\"}' | jq -c '{match_count, lists: .lists_screened}'"],"expect":"step 1 prints the control with canary.last_flip within the last hour, numeric canary.delivered_after_s at or under 900, canary.within_budget true and lists_fresh true; step 2 prints exactly the nine real lists; step 3 prints invalid_request_error/invalid_parameter/lists; step 4 prints match_count 0 with no canary in lists_screened","audit":"every flip is an ordinary list import: publish_info, designation changelog, archive, delta outbox row, reverse-screen job and alert all persist in the Noble-held tenant under the standard retention rules, and the receiver's canary-delivered log line carries the flip UID and trace id","regression":"TestCanaryInvisibleToCustomers, TestInternalSourceTenantFilter, TestCanaryReceiverMeasuresFlip, TestFlipDiff"}},{"date":"2026-11-02","category":"fixed","summary":"Every screening states its coverage, and a degraded screening fails closed: the coverage object on every /v1/screen response, 503 coverage_degraded by default, allow_degraded_coverage to elect the reduced answer on the record","details":"A sanctions screening must state the coverage it was computed over. Every /v1/screen response now carries coverage: {degraded, requested, expected, screened, issues[]} — on healthy and degraded responses alike, so a consumer never infers coverage from absence. Two conditions make a screening degraded: a screened list whose ingestion-freshness verdict is stale (the same verdict GET /v1/lists and /status already report), and lists_screened being narrower than the set the screening should have covered. The contract already promised fail-closed for an unconfirmable list; that promise now also covers staleness and the subset case: a degraded live screening is refused with 503 coverage_degraded, naming the affected lists and the opt-in, instead of answering as though nothing were wrong — an all-clear computed over missing or stale coverage is indistinguishable from a real one, which is exactly why it must not exist. Passing allow_degraded_coverage: true receives the reduced answer with coverage.degraded: true and one issue per affected list, and the election is stored with the request on the audit record. as_of screenings read history and are not gated on today's ingestion freshness. The coverage object is additive and the refusal enforces an already-documented promise, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Abu Abbas\"}' | jq '{degraded: .coverage.degraded, screened: (.coverage.screened|length), expected: (.coverage.expected|length), issues: .coverage.issues}'","STALE=$(curl -s https://api.noblesight.io/v1/lists | jq -r '.lists[] | select(.status==\"stale\") | .list' | head -1); echo \"stale list: ${STALE:-none}\"","[ -n \"$STALE\" ] \u0026\u0026 curl -s -o /tmp/r.json -w '%{http_code}' -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d \"{\\\"name\\\":\\\"Roman Abramovich\\\",\\\"lists\\\":[\\\"$STALE\\\"]}\"; jq '{code, message}' /tmp/r.json"],"expect":"step 1 prints coverage on every response: degraded false with equal screened/expected counts and issues [] while all lists are fresh; when any list reports stale on GET /v1/lists, step 3 against that list returns HTTP 503 with code coverage_degraded naming the list and allow_degraded_coverage, and repeating it with allow_degraded_coverage: true returns 200 with coverage.degraded true and a list_stale issue carrying last_checked","audit":"the stored screening record carries allow_degraded_coverage with the rest of the request, so a degraded answer is always a recorded election; GET /v1/lists reports the same freshness verdict the refusal cites","regression":"TestBuildScreenCoverage, TestStaleRequestedLists, TestScreenCoverageContract"}},{"date":"2026-11-02","category":"added","summary":"Monitoring latency is measured and published: GET /v1/lists traces every hop of each list's most recent change, and GET /v1/account/coverage gains the monitoring_latency control","details":"Nothing measured how long a sanctions-list change took to reach a customer. Three timestamps now exist and two surfaces compute the durations. Each list row on GET /v1/lists carries last_change: the most recent content change traced publisher → detected → imported → queued → re-screened → first alert → first webhook delivery, with the duration of every hop in whole seconds. The publisher's own instant (its HTTP Last-Modified) is recorded at import as source_last_modified; where a publisher exposes none, source_time says \"unknown\" and no publisher-side duration is invented. The alert and delivery hops are the calling tenant's own — send your API key to the (still public) endpoint and they scope to your tenant, with null_reason stating why they are null when they are. GET /v1/account/coverage gains an eighth control, monitoring_latency: every list change in the window, the p50/p95/max of detection-to-re-screened (basis rescreened), or detection-to-your-first-webhook-delivery where your account had deliveries (basis delivered), against the 900-second notify budget. Additive, so no version is minted.","endpoint":"/v1/lists","verification":{"steps":["curl -s https://api.noblesight.io/v1/lists -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.lists[]|{list, c: .last_change|{source_time, detected_after_s, queued_after_s, rescreened_after_s, delivered_after_s, null_reason}}'","curl -s https://api.noblesight.io/v1/account/coverage -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.controls[]|select(.id==\"monitoring_latency\")|{id, status, latency}'"],"expect":"step 1 prints a last_change object for every list that has recorded a change, with source_time in [\"publisher\",\"unknown\"], numeric queued_after_s and rescreened_after_s once its jobs completed, and detected_after_s only under source_time \"publisher\"; step 2 prints the monitoring_latency control with latency.window_days, latency.changes, latency.budget_s 900, numeric p50_s/p95_s/max_s, boolean within_budget and basis in [\"rescreened\",\"delivered\"]","audit":"each duration is the difference of the two stored instants the response itself names; the detected instant is the delta outbox row written in the import transaction, the same instant the coverage control measures from","regression":"TestComposeLastChange, TestNearestRank, TestFeedSourceLastModified"}},{"date":"2026-11-01","category":"fixed","summary":"GET /v1/portfolio?q= and GET /v1/portfolio/stats?q= now match Confidential Portfolio (encrypted) names: searching an encrypted portfolio by name finds the enrolled entity instead of returning an empty page","details":"On a Confidential Portfolio account, portfolio names are encrypted at rest under the account's dedicated key, so the database cannot see them. The q name filter was applied in the database, which silently excluded every encrypted entity: searching for an enrolled, monitored name returned an empty result that read as \"not monitored\". The filter now matches encrypted names in memory after decryption — the same way monitoring already screens them — so q finds an encrypted entity exactly as it finds a plaintext one, with unchanged cursor pagination. GET /v1/portfolio/stats applies the same filter the same way, so a filtered total always agrees with the filtered list. Unfiltered listing, score-band filters, and plaintext accounts are unchanged.","endpoint":"/v1/portfolio","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/portfolio?limit=1\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -r '.data[0].full_name'","curl -s \"https://api.noblesight.io/v1/portfolio?q=\u003cthat full_name, URL-encoded\u003e\u0026limit=100\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '[.data[].full_name]'","curl -s \"https://api.noblesight.io/v1/portfolio/stats?q=\u003cthat full_name, URL-encoded\u003e\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq .total"],"expect":"on any account with a monitored entity — Confidential Portfolio or plaintext — step 2 lists the entity named in step 1, and step 3 returns a total of at least 1 that equals the number of names step 2 lists","audit":"GET /v1/portfolio unfiltered lists the same entity with the same id; on a Confidential Portfolio account, GET /v1/account shows features.confidential_portfolio true while the search still finds the name","regression":"TestConfidentialPortfolioSearch, TestConfidentialPortfolioSearchPagination"}},{"date":"2026-10-30","category":"added","summary":"Confidential Portfolio (BYOK) key provisioning is live: an entitled account's POST /v1/portfolio/key provisions its dedicated encryption key instead of answering 503","details":"The Enterprise Confidential Portfolio add-on encrypts an account's portfolio names at rest under a key dedicated to that account, decrypted only in memory while screening runs. The /v1/portfolio/key endpoint shipped earlier, but the key service behind it was not configured in production, so every call from an entitled account answered 503 billing_not_enabled. The service is now configured: POST /v1/portfolio/key returns 201 with status active on first provisioning and is idempotent after (200 with the same status), GET returns the key's status and created_at, and DELETE revokes it — revocation is a final crypto-shred, after which the account's encrypted portfolio names are unreadable by anyone, including Noble, and the key cannot be re-provisioned. Accounts without the add-on continue to receive 403 tier_insufficient. Additive, so no version is minted.","endpoint":"/v1/portfolio/key","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/portfolio/key -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{status, created_at: (.created_at != null)}'","curl -s https://api.noblesight.io/v1/portfolio/key -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{status}'"],"expect":"with the confidential_portfolio add-on, step 1 returns {\"status\":\"active\",\"created_at\":true} (HTTP 201 the first time, 200 on every call after — provisioning is idempotent) and step 2 returns {\"status\":\"active\"}; without the add-on both return the 403 tier_insufficient error, and no entitled account sees a 503 billing_not_enabled","audit":"GET /v1/account shows features.confidential_portfolio true for the entitled account; GET /v1/portfolio/key reports the same status and created_at on every later call","regression":"TestPortfolioKeyEndpoints, TestRequireConfidentialPortfolioGate, TestConfidentialPortfolioEncryptsAtRest, TestConfidentialPortfolioReverseScreen, TestConfidentialPortfolioCryptoShredFailsClosed"}},{"date":"2026-10-30","category":"added","summary":"GET /v1/alerts accepts portfolio_entity_id: list the alerts monitoring has raised on one monitored portfolio record","details":"Every portfolio monitoring alert already names the monitored record it was raised on in portfolio_entity_id. The list endpoint now accepts that id as a filter, so one request answers \"what has monitoring found for this customer\" — pass the id GET /v1/portfolio returns (and the alert itself carries) and receive only that record's alerts, instead of paging the whole queue and filtering client-side. The filter combines with every existing one (status, source, score range, assignee, date range) and follows the same strictness rule: a malformed or non-positive value is a 400 invalid_parameter naming portfolio_entity_id, never a silently unfiltered 200. The noble CLI and MCP list_alerts tool accept the same filter. Additive, so no version is minted.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/portfolio?limit=1\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -r '.data[0].id'","curl -s \"https://api.noblesight.io/v1/alerts?portfolio_entity_id=\u003cthe id from step 1\u003e\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '{count: (.data | length), ids: [.data[].portfolio_entity_id] | unique}'","curl -s \"https://api.noblesight.io/v1/alerts?portfolio_entity_id=abc\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.error | {code, param}'"],"expect":"step 2 returns only alerts whose portfolio_entity_id equals the id you passed — ids is [] (no alerts for that record, also correct) or exactly [\u003cthat id\u003e]; step 3 returns {\"code\":\"invalid_parameter\",\"param\":\"portfolio_entity_id\"}","audit":"GET /v1/alerts/{id} for any alert step 2 lists shows the same portfolio_entity_id, and the monitored record with that id appears in your GET /v1/portfolio listing","regression":"TestListAlertsFilterByPortfolioEntity, TestListAlertsRejectsBadPortfolioEntityID"}},{"date":"2026-10-30","category":"fixed","summary":"Portfolio monitoring now evaluates every monitored customer against every listed company and vessel on the Canada list and every EU entity or alias published as a single name; engine 2.5.1 re-screens the portfolio against those records.","details":"Some lists publish a party as one name rather than a first and last name: Canada's companies and vessels, most EU entities, and most EU aliases. POST /v1/screen found those records, but portfolio monitoring had not evaluated monitored customers against them, at a list publish or at an engine re-screen. Engine 2.5.1 names every record the way screening does, and a catch-up re-screen evaluates every monitored customer against the Canada and EU lists. Each alert it raises carries trigger.kind engine_change, trigger.engine_version 2.5.1 and the trigger.publish_date of the list publish the record came from, so an analyst can see the designation predates the alert. A pair an analyst already closed as a true match is not re-alerted, and one closed as a false positive is re-alerted only on a higher score, as for every engine re-screen. Scores are unchanged: a monitoring hit remains the POST /v1/screen hit for the same pair.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/alerts?source=monitoring\u0026limit=200\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '[.data[] | select(.trigger.engine_version == \"2.5.1\")] | map({id, portfolio_entity_id, score, list: .trigger.list, kind: .trigger.kind, publish_date: .trigger.publish_date, uid: .match_payload[0].uid})'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"\u003cthe monitored customer name behind one of those alerts\u003e\"}' | jq -c '[.matches[] | select(.uid == \"\u003cthe uid from step 1\u003e\")] | first | {uid, score}'"],"expect":"every alert the first line lists has kind engine_change and list canada or eu, a uid prefixed CA- or EU-, and a publish_date on or before the day the alert opened; the second line returns the same uid with the same score the alert carries. A portfolio with no customer near a Canada company or vessel or an EU entity lists nothing, which is also correct.","audit":"GET /v1/alerts/{id} for any alert the first line lists shows the same trigger and match_payload; GET /v1/export?trace_id=\u003cits trace_id\u003e returns the screening record behind it with engine_version 2.5.1.","regression":"TestProbeNameFallsBackToWholeName, TestExtractScreenNamesUsesWholeNames, TestChunkChangesReportsDroppedRecords, TestActiveNamesNamesEveryLiveRecord, TestActiveNamesCanadaEntityOnlyRecord, TestEntitySnapshot_ToScreeningChange_WholeNameAlias"}},{"date":"2026-10-29","category":"added","summary":"A portfolio monitoring alert now says why it opened: trigger names the event, the list publish that carried it and when your customer was evaluated, and portfolio_entity_id names the monitored record","details":"A monitoring alert opens on a customer who may have screened clean for months, and the analyst has to explain why it opened that day. The alert stored the list and publish but returned neither, and nothing told a list event apart from a re-screen Noble ran itself. GET /v1/alerts and GET /v1/alerts/{id} now return trigger on every monitoring alert raised from today. kind is list_change (the list published a change to the listed party), engine_change (a re-screen after a change to Noble's matching rules, with engine_version), operator_reprobe (Noble re-screened the listed party on request; no list changed) or variation_added (Noble added spelling or transliteration variations of the party's names; no list changed). list, publish_id and publish_date name the publish as GET /v1/lists reports it, imported_at is when Noble ingested it, screened_at is when the customer was evaluated, and a list_change carries change_type added or modified. A later event that matches an open alert does not change its trigger: it is added to the timeline as a list_event activity entry carrying its own trigger. The same trigger is on the screening record behind the alert (GET /v1/export, response.trigger) and in the alert.created webhook payload. A monitoring alert also carries portfolio_entity_id, the id GET /v1/portfolio returns for the monitored record. api, batch and ownership alerts, and monitoring alerts raised before today, carry neither. Additive, so no version is minted.","endpoint":"/v1/alerts","verification":{"steps":["A=$(curl -s \"https://api.noblesight.io/v1/alerts?source=monitoring\u0026limit=100\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '[.data[] | select(.trigger)] | last')","echo \"$A\" | jq -c '{kind: .trigger.kind, list: .trigger.list, publish_id: .trigger.publish_id, publish_date: .trigger.publish_date, screened_at: .trigger.screened_at, portfolio_entity_id}'","curl -s https://api.noblesight.io/v1/lists | jq -c '.lists[] | {list, publish_id, publish_date}'"],"expect":"the second line prints a kind of list_change, engine_change, operator_reprobe or variation_added, the list, a publish_id with its publish_date, an RFC 3339 screened_at and a positive portfolio_entity_id; for a list_change on the current publish, publish_id and publish_date equal that list's entry in the third command","audit":"GET /v1/export?trace_id=\u003cthe screening's trace_id\u003e returns the screening record behind the alert, whose response.trigger equals the alert's trigger under jq -S; each list_event entry in GET /v1/alerts/{id} activity carries the trigger of the event it records","regression":"TestMonitoringAlertTrigger, TestJobTriggerKind, TestTriggerForRecord, TestGetAlertDecodesTrigger"}},{"date":"2026-10-29","category":"fixed","summary":"Portfolio monitoring no longer skips a Confidential Portfolio customer when the key service fails briefly: the job waits and retries, and a customer skipped for a disabled or destroyed key is counted on GET /status","details":"Each sanctions-list change is screened against every monitored portfolio, and a Confidential Portfolio customer's entities are decrypted with that customer's key for every job. Any failure to decrypt was treated as a destroyed key: the customer's entities were left out of that change and the job still finished, so a brief key service failure meant the change was never screened against that customer's portfolio. A temporary failure now fails the job, which retries after 1, 5 and then 15 minutes and screens every customer once the key service answers, so no change is screened for some customers and not others; while it waits it shows in monitoring.reverse_pending. If the key service still fails for one customer's key on the last retry, that customer alone is skipped and every other portfolio is screened, so one customer's key cannot hold up monitoring for everyone. A customer whose key is disabled or destroyed, or whose stored record does not decrypt, is skipped on purpose and not retried. GET /status now counts the customers the most recently finished job skipped under monitoring.reverse_encrypted_skipped, with reverse_encrypted_skipped_oldest_s the seconds the longest-skipped has been skipped without a break. A customer that revoked its key with DELETE /v1/portfolio/key has nothing left to screen and is not counted. Two fields are added to the monitoring object; no field changed, so no version is minted.","endpoint":"/status","verification":{"steps":["curl -s https://api.noblesight.io/status | jq -c '.monitoring | with_entries(select(.key | test(\"encrypted\")))'"],"expect":"prints {\"reverse_encrypted_skipped\":0,\"reverse_encrypted_skipped_oldest_s\":0}: no Confidential Portfolio customer was skipped by the most recent job","audit":"Each skipped customer is recorded on the job with the reason, the number of entities and, for a record that does not decrypt, its portfolio entity id, and logged once per job; a temporary failure is recorded as the job's retry, and becomes a skip only on the last one","regression":"TestReverseScreenJobEncryptedSkips, TestPipelineStatusEncryptedSkipsByFinish, TestUnwrapError, TestMergeEncryptedSkips, TestAddSkip"}},{"date":"2026-10-29","category":"fixed","summary":"POST and PUT /v1/portfolio screen the entity before writing it and return that screening; a name already on a list is alerted at enrollment","details":"Adding an entity with POST /v1/portfolio, or changing one with PUT, wrote it and screened nothing, though the docs said it was screened immediately. A name already on a sanctions list raised no alert until its listed record next changed, which may be never. Both now run the screening POST /v1/screen runs, against every list, with the entity's full_name, date_of_birth and country, and write the entity only after it succeeds. The screening is recorded like any other (GET /v1/export?trace_id= returns it, and it replays) and raises an alert through the same threshold and dedup. The response carries it under screening: trace_id, screening_id, screened_at, match_count, top_match (the /v1/screen match, null when nothing matched), alert_id (null when none) and alert_reason. The entity's highest_score, average_score and last_screened_at are that screening's. When the screening cannot run (503 coverage_unavailable or 504 screening_timeout), nothing is written and the attempt is recorded as a failed screening; retry the request. date_of_birth must now be YYYY-MM-DD and country a two-letter code (either case, stored upper-case); anything else is a 400 invalid_parameter rather than a 500 or a value monitoring could not compare. Additive, so no version is minted.","endpoint":"/v1/portfolio","verification":{"steps":["E=$(curl -s -X POST https://api.noblesight.io/v1/portfolio -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"Joaquin Guzman Loera\"}')","echo \"$E\" | jq -c '{id, trace: .screening.trace_id, uid: .screening.top_match.uid, alert: .screening.alert_id}'","curl -s -X PUT \"https://api.noblesight.io/v1/portfolio?id=$(echo \"$E\" | jq -r .id)\" -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"Joaquin Guzman Loera\",\"country\":\"MX\"}' | jq -c '{country, trace: .screening.trace_id}'"],"expect":"on a key with the portfolio add-on, the first line prints a non-null trace, uid \"6861\" and a non-null alert (the alert raised now, or the open one it was deduplicated into); the second prints country \"MX\" and a different non-null trace","audit":"GET /v1/export?trace_id={screening.trace_id} returns the recorded screening, whose response carries enrollment {status, portfolio_entity_id} naming the entity written; GET /v1/alerts/{alert_id} carries the /v1/screen match for uid 6861 in match_payload","regression":"TestPortfolioWriteScreens, TestAddToPortfolioDecodesScreening"}},{"date":"2026-10-29","category":"fixed","summary":"A name enrolled by POST /v1/screen is monitored with the date of birth and country it was screened with; a screening with no name enrolls nothing and reports enrollment.status no_name","details":"With the portfolio add-on, a name you screen is enrolled for continuous monitoring. Enrollment stored only the name: the date_of_birth and country on the request were dropped, so when a list later changed, the customer was compared without them, scoring higher than the screening that enrolled it did. Enrollment now stores the date_of_birth (when it is YYYY-MM-DD, the form screening compares) and the country (a two-letter code, upper-cased), so a monitoring hit equals the /v1/screen hit for the request that enrolled the name. A later screening of the same name without them keeps the stored values; one with new values replaces them. A screening with no name, such as an identifier screened alone, used to enroll an empty name and report enrolled, though monitoring compares names, so nothing could ever match it. It now enrolls nothing and reports the new enrollment.status no_name: identifiers are not monitored. Empty-name portfolio entries left by earlier screenings are removed. Additive, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Enroll Attr Probe 2026-10-29\",\"date_of_birth\":\"1970-01-01\",\"country\":\"MX\"}' | jq -c .enrollment","curl -s \"https://api.noblesight.io/v1/portfolio?q=Enroll%20Attr%20Probe%202026-10-29\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.data[0] | {date_of_birth, country}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"id_number\":\"ENROLL-PROBE-0001\"}' | jq -r .enrollment.status"],"expect":"on a key with the portfolio add-on: the first prints {\"status\":\"enrolled\",\"portfolio_entity_id\":N} (already_enrolled on a repeat), the second prints {\"date_of_birth\":\"1970-01-01\",\"country\":\"MX\"}, and the third prints no_name","audit":"GET /v1/export?trace_id={trace_id} of the identifier-only screening returns a record whose response carries enrollment {\"status\":\"no_name\"}","regression":"TestScreenEnrollmentReported, TestEnrollmentAttrs, TestEnrollScreenedNeedsNoDatabase, TestEnrollmentStatusesMatchMonitor"}},{"date":"2026-10-29","category":"fixed","summary":"POST /v1/batch enrolls a screened name only with the portfolio add-on, and every screened item reports enrollment as POST /v1/screen does","details":"A batch enrolled every name it screened into portfolio monitoring whether or not the key held the portfolio add-on, and a failed enrollment was only logged on Noble's side, so a completed batch could leave names unmonitored with nothing in the result to say so. Each screened item now enrolls exactly as POST /v1/screen does and carries enrollment: {status, portfolio_entity_id}. With the add-on the name is enrolled with the item's date_of_birth and country (enrolled or already_enrolled); without it the status is not_entitled and nothing is enrolled; a failure is failed, with an enrollment_failed warning on the item. Items screened before this carry no enrollment field. Names a batch enrolled earlier for a key without the add-on stay enrolled. Additive, so no version is minted.","endpoint":"/v1/batch","verification":{"steps":["J=$(curl -s -X POST https://api.noblesight.io/v1/batch -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"names\":[{\"full_name\":\"Batch Enroll Probe 2026-10-29\",\"date_of_birth\":\"1970-01-01\",\"country\":\"MX\"}]}' | jq -r .id)","sleep 30; curl -s \"https://api.noblesight.io/v1/batch/status?id=$J\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -c '.items[0].enrollment'"],"expect":"prints {\"status\":\"enrolled\",\"portfolio_entity_id\":N} (already_enrolled on a repeat) on a key with the portfolio add-on, or {\"status\":\"not_entitled\"} on a key without it; never no enrollment field on an item screened from now on","audit":"GET /v1/export?trace_id={the batch's trace id} returns each item's batch screening record, whose response carries the same enrollment object","regression":"TestBatchEnrollmentReported, TestBatchDecodesEnrollment, TestEnrollmentStatusesMatchMonitor"}},{"date":"2026-10-28","category":"fixed","summary":"A portfolio monitoring alert now carries exactly the hit POST /v1/screen returns for the same customer and listed party: same score, match sources, signals, programs and identifiers; engine_version 2.5.0","details":"When a sanctions list changes, Noble compares the changed entry with your monitored portfolio. That comparison used to be its own scorer, which agreed with /v1/screen on the score of most pairs but not on everything else: it skipped the OFAC FAQ 5 partial-name control and the date-of-birth and country comparisons, reported every hit as coming through a pathway called reverse, carried no programs, addresses or identifiers, and stored the hit with different field names than /v1/screen. Now both directions decide a pair with one function. A monitoring alert's hit is the match /v1/screen returns when that customer's name is screened, for the same listed party, field for field: score, match_sources (with the real pathway, such as ofac_sdn or ofac_aka, and the matched name), dismissal_signals, programs, prohibition_type, addresses, ids, date_of_birth, place_of_birth and nationalities. The date of birth and country stored on a portfolio entity count exactly as they would if sent on a screen, and are absent when you never set them. A changed entry is compared under every name Noble holds for it, not only the names the change carried. WHAT YOU WILL SEE: engine_version \"2.5.0\"; on a monitoring alert raised from now on, match_payload is an array of /v1/screen match objects in snake_case (uid, score, match_sources, dismissal_signals), as on a screening alert; batch alerts store the same shape and now carry programs and identifiers too. Alerts raised before this keep the shape they were stored in. A monitoring alert can now score lower than before where the partial-name control or a date of birth or country rules a pair out, exactly as /v1/screen does. Because this changes what monitoring finds, monitored portfolios are re-screened; a pair you closed as a false positive is raised again only if it now scores higher than when you closed it, and one you closed as a true match is never raised again. match_payload's contents were never part of the typed contract, so no version is minted.","endpoint":"/v1/alerts","verification":{"steps":["ALERT=$(curl -s \"https://api.noblesight.io/v1/alerts?source=monitoring\u0026limit=100\" -H \"X-API-Key: $NOBLE_API_KEY\" | jq -r '[.data[] | select(.created_at \u003e= \"2026-10-29\")][0].id')","curl -s https://api.noblesight.io/v1/alerts/$ALERT -H \"X-API-Key: $NOBLE_API_KEY\" \u003e alert.json","jq -n --slurpfile a alert.json '{name: $a[0].entity.full_name, limit: 100} + ($a[0].entity | {date_of_birth, country} | with_entries(select(.value != null and .value != \"\")))' | curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d @- \u003e screen.json","diff \u003c(jq -S --arg u \"$(jq -r '.match_payload[0].uid' alert.json)\" '.matches[] | select(.uid == $u)' screen.json) \u003c(jq -S '.match_payload[0]' alert.json) \u0026\u0026 echo identical"],"expect":"prints identical: the monitoring alert's hit and the /v1/screen match for the same listed party are the same object; and jq '[.match_payload[].match_sources[].source] | index(\"reverse\")' alert.json prints null","audit":"GET /v1/export?channel=monitoring returns the monitoring evidence record for the alert, whose matches carry the same hit beside the portfolio entity it was decided for, with engine_version 2.5.0, match_engine index and the list_versions it was decided against","regression":"TestReverseHitEqualsScreen, TestReverseRetrievalSuperset, TestDecideEqualsScreen, TestDecideRunsStep3, TestScreenRecordEqualsScreen, TestFinalizeHitsDecidesEachHitAlone, TestRealertReason"}},{"date":"2026-10-27","category":"added","summary":"POST /v1/screen reports what it did to your portfolio: enrollment.status is enrolled, already_enrolled, not_entitled or failed","details":"With the portfolio add-on, every name you screen is also enrolled for continuous monitoring. Enrollment used to happen after the response, and a failure was only logged on Noble's side, so a screening could return a complete-looking 200 while the name was never monitored. Enrollment now completes before the response, which carries enrollment: {status, portfolio_entity_id}. enrolled means this screening added the name; already_enrolled means it was already monitored; not_entitled means the key has no portfolio add-on; failed means the name is not monitored, and the response also carries an enrollment_failed warning: screen it again or add it with POST /v1/portfolio. portfolio_entity_id is the id GET /v1/portfolio returns. An as_of screening never enrolls and has no enrollment field. The outcome is recorded with the screening, so an export shows what you saw. Additive, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Enrollment Probe 2026-10-27\"}' | jq -c .enrollment"],"expect":"prints {\"status\":\"enrolled\",\"portfolio_entity_id\":N} the first time (already_enrolled with the same id on a repeat) on a key with the portfolio add-on, or {\"status\":\"not_entitled\"} on a key without it; never no enrollment field on a present-day screening","audit":"GET /v1/export?trace_id={trace_id} returns the recorded screening, whose response carries the same enrollment object","regression":"TestScreenEnrollmentReported, TestScreenDecodesEnrollment"}},{"date":"2026-10-27","category":"fixed","summary":"Every API key belongs to its account's tenant, so a screening can no longer raise an alert the same key cannot read","details":"Alerts, portfolio monitoring and webhooks are scoped to the tenant an account belongs to. A few keys issued before tenants existed carried none: such a key could screen, enroll names and raise alerts, while every alert endpoint refused it with 403 tenant_not_provisioned and no webhook carried those alerts. Each such account now has its own tenant, and its keys, alerts and other records were moved onto it, so the alerts they raised are readable by the keys that raised them. A key can no longer exist without a tenant, and POST /v1/screen, POST /v1/batch and POST /v1/portfolio refuse one with 403 tenant_not_provisioned, as the alert endpoints already did, rather than create anything its own key cannot see. No current key is affected. No response shape changed, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["ALERT=$(curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $NOBLE_API_KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq -r .alert_id)","curl -s -o /dev/null -w '%{http_code}\\n' https://api.noblesight.io/v1/alerts/$ALERT -H \"X-API-Key: $NOBLE_API_KEY\""],"expect":"ALERT is a positive integer and the second command prints 200: the alert a screening raises is readable by the key that screened","audit":"Each record moved onto a provisioned tenant is listed, with the tenant it moved to, in Noble's tenant merge record","regression":"TestTenantForEveryKeyMigration, TestTenantForEveryKeyMigrationRefusesTakenTenant"}},{"date":"2026-10-27","category":"fixed","summary":"A sanctions publisher's brief outage no longer delays list updates by a whole poll interval: Noble retries the download within seconds","details":"Noble checks each sanctions list for changes every 5 minutes (OFAC's SDN Advanced export hourly). A check that failed for any reason, such as a publisher resetting the connection, answering with a server error or an HTML maintenance page, or responding slowly, was abandoned until the next check, so a list change published during a run of failures went unseen until the publisher recovered; in the 30 days to 2026-10-25 that happened 1,079 times, the longest run 19 hours on the EU list. A transient failure is now retried up to three more times in the same check, with waits totalling at most 90 seconds between attempts, growing each time and honouring the publisher's Retry-After; no retry starts once the next check is due. A failure that would recur, such as a missing file or a list larger than the size limit, is not retried. Downloads are given time in proportion to the list's size instead of one 2-minute limit for every list, and a download that stops sending data for a minute is abandoned and retried, and every request identifies Noble with a contact address. No response shape changed, so no version is minted.","endpoint":"/v1/lists","verification":{"steps":["curl -s https://api.noblesight.io/v1/lists | jq -c '[.lists[] | select(now - (.last_checked | fromdateiso8601) \u003e .fresh_within_s) | .list]'"],"expect":"prints []: every list was checked within its freshness window","audit":"Each failed check is logged with the attempts made and its error class, and each check that succeeded only after a retry is logged as recovered, so retries are distinguishable from hard failures","regression":"TestDownloadRetries, TestRetryStopsAtBudgetAndDeadline, TestClassify, TestTickRetry, TestFeedOpenRejectsHTML, TestFeedCheckRejectsHTML, TestFeedBodyTimeout, TestFeedStallTimeout, TestBodyTimeoutScalesWithFeed"}},{"date":"2026-10-27","category":"fixed","summary":"Portfolio monitoring now finishes every sanctions-list change: a reverse-screen job that fails retries itself, and a large change is split into jobs instead of held","details":"Each sanctions-list change is screened against every monitored portfolio by a reverse-screen job. A job that hit an error, such as a database restart mid-run, stopped as failed until an operator stepped in, and a change of more than 5,000 names was held for review, so in both cases the change was not screened against portfolios until someone acted. A failed job now retries on its own after 1, 5 and then 15 minutes, resuming after the names it already screened; only a job whose third retry also fails stays failed, and Noble is alarmed on it. A change over 5,000 names is split into jobs of at most 5,000 and every job runs. A crashed job is resumed once it has made no progress for ten minutes, the same bound reverse_processing_over_heartbeat on GET /status counts. The changes held before this fix are screened. No response shape changed, so no version is minted.","endpoint":"/status","verification":{"steps":["curl -s https://api.noblesight.io/status | jq -c '.monitoring | {reverse_failed, reverse_skipped, reverse_processing_over_heartbeat}'"],"expect":"prints {\"reverse_failed\":0,\"reverse_skipped\":0,\"reverse_processing_over_heartbeat\":0}; a transient failure shows in reverse_pending while its retry waits, never in reverse_failed","audit":"GET /v1/alerts?source=monitoring lists the alerts the previously held changes raised, each citing the list publish that produced it","regression":"TestReverseScreenJobRetry, TestReverseScreenJobOversizeChunked, TestReverseScreenJobReclaimOnHeartbeat, TestAdminRequeueReverseJob, TestRetryDelay"}},{"date":"2026-10-27","category":"fixed","summary":"The first as_of screening at a list version Noble has not prepared yet is now answered, not refused with 503 coverage_unavailable","details":"An as_of screening, or a replay of an older record, is scored against the sanctions lists exactly as they stood at that instant, and a server prepares that list version the first time it is asked for. Preparing one took about as long as a screening is allowed to wait, so the request that asked for it was often answered 503 coverage_unavailable while the version finished in the background, and only a retry was answered. Preparation is now faster, and the first request for a list version is answered within the screening deadline, in about 5 to 6 seconds; later requests at that version, and at any as_of between the same two list updates, are answered at normal speed. A server prepares one list version at a time, so concurrent requests needing other versions are still answered 503 coverage_unavailable with a Retry-After header (5 seconds), and a retry after it is answered. No response shape changed, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["for d in 2026-06-0{1..9} 2026-09-0{1..9}; do curl -s -o /dev/null -w \"%{http_code} %header{retry-after} %{time_total} $d\\n\" -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"name\\\":\\\"Vladimir Putin\\\",\\\"as_of\\\":\\\"${d}T00:00:00Z\\\"}\" \u0026 done; wait","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"as_of\":\"2026-09-07T00:00:00Z\"}' | jq -r '.matches[0].uid'"],"expect":"step 1 prints 18 lines, each 200 or 503 and never 502, 504 or an empty reply; every 503 line shows a Retry-After of 5; at least one line is 200 in under 10 seconds. Step 2 prints 35096","audit":"each answered as_of screening reports in list_versions the list version it was scored against, and a replay of it is scored against that same version","regression":"TestBuildIndexAtMatchesSequential, TestCutIndexesCapConcurrentBuilds, TestCutIndexesShareOneVector, TestIndexUnavailableScopedToScreening"}},{"date":"2026-10-27","category":"added","summary":"DELETE /v1/account/keys/{key_prefix}?confirm={key_prefix} revokes one of your keys; a revoked key is answered 401 api_key_revoked","details":"A key could not be revoked: one that leaked into a log or a repository stayed valid, and the old key after a rotation could not be retired. A key can now revoke itself, or a key on the account issued before it, by its first 16 characters; confirm must repeat them so a stray request revokes nothing, and an optional reason is recorded. After a rotation the new key retires the old one; a key cannot revoke a key issued after it (403 api_key_revoke_forbidden), so a leaked old key cannot revoke its replacement. Revocation is permanent and applies on the key's very next request: a revoked key gets 401 with code api_key_revoked, distinct from api_key_invalid for a key Noble does not recognize (it was 403 api_key_invalid). The key's screenings, alerts and usage stay on the account. The call never counts against quota, so a key that has used up its quota can still be revoked. Repeating it returns the original revocation with already_revoked true; revoking the last active key is allowed, with active_keys_remaining 0 and a warning. A key prefix that is not on the caller's account is 404, whether or not it exists elsewhere. Revocation stops the credential, not what it did: after revoking a leaked key, check GET /v1/webhooks for endpoints you did not register. Noble can revoke a key for you; contact support@noblesight.io.","endpoint":"/v1/account/keys","verification":{"steps":["curl -s -o /dev/null -w '%{http_code}\\n' https://api.noblesight.io/v1/usage -H \"X-API-Key: $A\"","curl -s -X DELETE \"https://api.noblesight.io/v1/account/keys/${A:0:16}?confirm=${A:0:16}\" -H \"X-API-Key: $B\" | jq -c '{key_prefix, status, revoked_at, active_keys_remaining}'","curl -s -w '\\n%{http_code}\\n' https://api.noblesight.io/v1/usage -H \"X-API-Key: $A\"","curl -s -o /dev/null -w '%{http_code}\\n' https://api.noblesight.io/v1/usage -H \"X-API-Key: $B\""],"expect":"with two keys of one account, $A and $B issued after it: step 1 prints 200; step 2 prints key_prefix equal to the first 16 characters of $A, status \"revoked\", an RFC 3339 revoked_at and active_keys_remaining 1; step 3 prints a body with \"code\":\"api_key_revoked\" and then 401; step 4 prints 200","audit":"each revocation is one append-only customers.api_key_revocation row, written with the status change, carrying the key, its prior status, who revoked it (the revoking key, or the operator), the reason and the request trace id","regression":"TestRevokeAccountKey, TestRevokeOwnKeySharingPrefix, TestRevokeByRevokedCaller, TestRevokeAccountKeyOtherClient, TestRevokeAccountKeyWhenQuotaExhausted, TestAdminRevokeKey"}},{"date":"2026-10-27","category":"fixed","summary":"Retiring a key no longer hides a customer's alerts, webhooks or team from its other keys: every key Noble issues to a customer shares one workspace","details":"Each key Noble issued, including a replacement key for a rotation, was given a workspace of its own, and alerts, webhooks, team members and alert activity were read from the workspace of the customer's oldest active key. While that key stayed active every key saw the same data; once it expired or was retired, the customer's other keys switched to a different workspace and stopped listing the alerts, webhooks and team members created before. Every key a customer holds now belongs to the customer's single workspace, so a rotation, an expiry or a retirement never changes what the remaining keys see. Customers who held more than one workspace have been consolidated onto the one their keys were reading, with alerts, their activity history, batch jobs, webhooks, pending webhook deliveries and team members carried over; the alert threshold in effect is unchanged. Where both workspaces held the same thing, the copy already visible is kept: a webhook URL registered on both keeps one active registration, so no alert is delivered twice, and a team member registered on both under one email keeps one registration. An open monitoring alert whose duplicate was already open in the consolidated workspace is represented by that duplicate. Alert listings can grow for customers whose earlier alerts had dropped out of view; response shapes are unchanged — no version change.","endpoint":"/v1/alerts","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $A\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq -r .alert_id \u003e /tmp/alert","Noble issues a second key $B for the same customer and retires $A (support@noblesight.io)","curl -s \"https://api.noblesight.io/v1/alerts/$(cat /tmp/alert)\" -H \"X-API-Key: $B\" | jq -r .id"],"expect":"step 1 writes a numeric alert id; step 3 prints the same id after $A is retired, where before it answered 404","audit":"every row moved during consolidation is recorded append-only with the workspace it came from, so each alert's original ownership stays recoverable; no screening fact, score or disposition changed","regression":"TestRetiredKeyKeepsClientAlerts, TestConcurrentFirstMintsShareTenant, TestTenantMergeMigration, TestTenantMergeMigrationRefusesSharedTenant"}},{"date":"2026-10-27","category":"security","summary":"POST /v1/keys treats client_id as a label: Noble assigns every new key its own client_id, the label followed by -\u003c8 hex\u003e, and returns it as client_id in the response","details":"client_id is the customer identity behind every grant, quota and record Noble keeps for an account, so it is now always assigned by Noble. The client_id you send at signup is a label: Noble stores the key under a client_id made from the label (cut to 91 bytes if longer) followed by a hyphen and 8 random hex characters, so every signup creates a new customer. The response is the same whether or not another customer used the label. CreateKeyResponse gains client_id, and GET /v1/account reports the same value. Adding a key to an existing account is done by Noble on request.","endpoint":"/v1/keys","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/keys -H 'Content-Type: application/json' -d '{\"client_id\":\"acme-bank\",\"name\":\"verify\",\"accept_terms\":true,\"invite_code\":\"'$INVITE'\"}' | tee /tmp/k.json | jq -c '{client_id, key_prefix}'","curl -s https://api.noblesight.io/v1/account -H \"X-API-Key: $(jq -r .key /tmp/k.json)\" | jq -r .client_id"],"expect":"step 1 returns HTTP 201 with a client_id matching ^acme-bank-[0-9a-f]{8}$, never exactly acme-bank, whether or not the label is already in use; step 2 prints the same client_id as step 1","audit":"the signup's api_key_minted log event records both the assigned client_id and the requested client_label, with the request trace id","regression":"TestSignupClientID, TestCreateKeySuccess, TestCreateKeyCannotJoinExistingClient, TestIdempotencyCreateKeyIPScope"}},{"date":"2026-10-25","category":"added","summary":"Enterprise quotas are set per contract: a custom monthly limit, or unlimited. GET /v1/usage and GET /v1/account report \"unlimited\": true, without limit or remaining, for a key with no quota","details":"Enterprise is sold as custom, 100,000+ screenings a month, but every Enterprise key was held to exactly 100,000 with no way to change it. Noble now applies the contract's number per customer: a monthly limit of any size, or no quota at all. The contract follows the customer, not the key: a replacement key Noble issues for a rotation carries the same quota, so a rotation never drops a customer to the tier default mid-month. An unlimited key is never refused with rate_limit_exceeded, and its responses carry no X-RateLimit-* headers because there is no limit to report. Its usage is still counted. GET /v1/usage and GET /v1/account gain an unlimited field, false for every limited key; for an unlimited key they omit limit and remaining rather than report a remaining of 0 that would read as exhausted. A client that decodes limit and remaining into plain integers will see 0 for both on an unlimited key and should check unlimited first. Limited keys are unchanged on the wire. A self-serve Stripe upgrade sets the purchased tier's quota. Every quota change, whether made by Noble, by checkout or on rotation, is recorded with its source, who made it, why, and when.","endpoint":"/v1/usage","verification":{"steps":["curl -s https://api.noblesight.io/v1/usage -H \"X-API-Key: $KEY\" | jq -c '{unlimited, limit, remaining, used, window}'","curl -s -D - -o /dev/null -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"vladimir putin\"}' | grep -iE '^(HTTP/|x-ratelimit)'","curl -s https://api.noblesight.io/v1/account -H \"X-API-Key: $KEY\" | jq -c '.usage | {unlimited, limit, remaining}'"],"expect":"on a limited key, step 1 prints unlimited false with integer limit and remaining, as before; on a key with no quota, step 1 prints {\"unlimited\":true,\"limit\":null,\"remaining\":null,\"used\":\u003cn\u003e,\"window\":\"month\"}, step 2 prints only an HTTP/2 200 status line (no X-RateLimit-* headers), while on a limited key it prints the 200 line plus X-RateLimit-Limit, X-RateLimit-Remaining and X-RateLimit-Window, and step 3 prints {\"unlimited\":true,\"limit\":null,\"remaining\":null}","audit":"each quota change is an append-only customers.api_key_quota_change row carrying the key, the monthly limit and unlimited flag before and after, the operator, the reason and the request trace id","regression":"TestAdminKeyQuota, TestAdminKeyQuotaClientKeys, TestAdminKeyQuotaRefusals, TestKeyQuota, TestCheckoutReplacesQuotaOnTheRecord, TestGetUsageDecodesUnlimited, TestGetUsageDecodesAllFields"}},{"date":"2026-10-25","category":"changed","summary":"GET /v1/lists measures freshness in poll intervals: a list is \"degraded\" after 30 minutes without a successful poll and \"stale\" after 2 hours, not 24 hours; each list states its bounds in fresh_within_s and stale_after_s","details":"Every list is now polled every five minutes; Belgium, the Netherlands and Australia moved from every 30 minutes. status was \"ok\" until a list had gone 24 hours without a successful poll, so an outage of most of a day on a list polled every five minutes read \"ok\" throughout. The verdict is now measured in the list's poll interval: \"ok\" while last_checked is within fresh_within_s (six intervals, 1800 seconds), \"degraded\" up to stale_after_s (24 intervals, 7200 seconds), and \"stale\" beyond it or when the last poll is unknown. \"degraded\" is a new value of status: several polls in a row have failed, Noble has been alarmed and is retrying, and screening continues against the snapshot the entry describes. The GET /status lists component reports \"degraded\" when any list is degraded or stale. ListFreshThreshold in the Go contract is deprecated and now equals the longest stale_after_s, 2 hours. Integrations that treat any status other than \"ok\" as unhealthy need no change; integrations that match on \"stale\" alone should add \"degraded\".","endpoint":"/v1/lists","verification":{"steps":["curl -s https://api.noblesight.io/v1/lists | jq -c '.lists[] | {list, status, last_checked, fresh_within_s, stale_after_s}'","curl -s https://api.noblesight.io/v1/lists | jq -c --arg now \"$(date -u +%s)\" '[.lists[] | {list, status, age_s: (($now|tonumber) - (.last_checked|fromdateiso8601)), fresh_within_s, stale_after_s} | select((.age_s \u003c= .fresh_within_s and .status != \"ok\") or (.age_s \u003e .fresh_within_s and .status == \"ok\"))]'","curl -s https://api.noblesight.io/status | jq -c '{status, lists: .components.lists.status}'"],"expect":"step 1 prints nine lists, each with fresh_within_s 1800, stale_after_s 7200 and status one of ok, degraded or stale, and last_checked within the last five minutes on a healthy deployment; step 2 prints [] (no list's status disagrees with its own age and bounds); step 3 prints lists \"operational\" when every list in step 1 is ok, and \"degraded\" otherwise","audit":"fresh_within_s and stale_after_s are served with every entry, so a recorded GET /v1/lists response carries the bounds its verdict was computed against; every screening response's list_versions names the snapshot it ran on","regression":"TestFreshness, TestFreshnessContractMatchesListhealth, TestStatusListsComponent, TestGetSanctionsListsDecodesEntries"}},{"date":"2026-10-25","category":"added","summary":"GET /status reports the portfolio-monitoring pipeline under monitoring: queued, stalled, failed and held reverse-screen jobs, webhook events waiting over 5 minutes and list changes unqueued over 15 minutes, each with the age of the oldest","details":"A sanctions-list change reaches a monitored portfolio in three steps: it is recorded as a change, screened against every monitored portfolio by a reverse-screen job, and each alert it raises is delivered to registered webhooks. None of those queues was visible. GET /status now carries a monitoring object with reverse_pending, reverse_processing_over_heartbeat (a job holding its claim with no progress for over ten minutes), reverse_failed, reverse_skipped (a change too large for one job, held for review), webhook_pending_over_5m, webhook_unattempted_over_5m and delta_pending_over_15m, each paired with an *_oldest_s age in seconds. Counts are deployment-wide and expose no tenant's data. The block is informational and does not change status; it is null when it cannot be read, never zero, and is refreshed at most every 15 seconds. delta_pending_over_15m is the same condition per-list monitoring \"delayed\" reports on GET /v1/lists. Noble is alarmed on each stall condition.","endpoint":"/status","verification":{"steps":["curl -s https://api.noblesight.io/status | jq -c '.monitoring'","curl -s https://api.noblesight.io/status | jq -c '.monitoring | [to_entries[] | select((.value|type) != \"number\")]'"],"expect":"step 1 prints an object (not null) with integer fields reverse_pending, reverse_processing_over_heartbeat, reverse_failed, reverse_skipped, webhook_pending_over_5m, webhook_unattempted_over_5m and delta_pending_over_15m, and an integer *_oldest_s beside each queue; step 2 prints []","audit":"delta_pending_over_15m above 0 coincides with monitoring \"delayed\" on the affected list in GET /v1/lists; GET /v1/alerts?source=monitoring shows the alerts a completed job raised","regression":"TestPipelineStatusCounts, TestStatusMonitoringBounds, TestStatusListsComponent"}},{"date":"2026-10-25","category":"fixed","summary":"The hosted MCP screening tool is renamed screen and accepts address, lists, as_of and ids, so every control the coverage report scores is reachable from an agent; its city description says a Crimean city raises the embargo advisory","details":"The tool formerly named screen_name is now screen: it screens a name or identifiers alone, mirrors POST /v1/screen and pairs with get_screening. screen_name is removed, not aliased, so an MCP client that pinned the old name must update its tool reference. Two parity gaps between the tool on mcp.noblesight.io and POST /v1/screen. First, the tool's city parameter was described as \"recorded for context; not screened and never raises an advisory on its own\", which has been false since 2026-10-24: location.city is scanned for a Crimean city or the word Crimea and raises the UA-43 or UA-40 advisory, and the tool did the same while telling the agent it could not. The description now says what the field does, including the possible status a lone town name under a contradicting country code reports. Second, the tool had no way to send address (the CoinList, Uphold and Poloniex control), lists (restrict screening to named lists), as_of (the Bank of America lookback control) or ids (identifier screening), so an MCP-driven integration could not exercise two of the seven controls GET /v1/account/coverage scores. All four are now parameters with the same semantics as REST; ids is an array of {number, type}, and name may be omitted when ids is given, so an agent can screen an IMO number, a wallet address or a BIC alone. A call with neither name nor ids is refused before any request is made. A test now binds every advertised screen parameter to a JSON field of the REST request type and asserts the city description, so the tool schema cannot drift from the contract silently again. The tool remains a client of the public API: what it sends is what /v1/screen receives and what the screening record holds.","endpoint":"/v1/screen","verification":{"steps":["curl -s https://mcp.noblesight.io/mcp -H \"Authorization: Bearer $KEY\" -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' -d '{\"jsonrpc\":\"2.0\",\"id\":1,\"method\":\"tools/list\"}' | sed -n 's/^data: //p' | jq -c '.result.tools[]|select(.name==\"screen\")|{params:(.inputSchema.properties|keys), city_never:(.inputSchema.properties.city.description|contains(\"never raises an advisory\")), city_crimea:(.inputSchema.properties.city.description|contains(\"Crimea\"))}'","curl -s https://mcp.noblesight.io/mcp -H \"Authorization: Bearer $KEY\" -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' -d '{\"jsonrpc\":\"2.0\",\"id\":2,\"method\":\"tools/call\",\"params\":{\"name\":\"screen\",\"arguments\":{\"name\":\"Zzyxx Cleanname\",\"country\":\"RU\",\"address\":\"ul. Kirova 1, Simferopol\"}}}' | sed -n 's/^data: //p' | jq -c '.result.structuredContent|{j:.jurisdiction_risk.jurisdiction, src:.jurisdiction_nexus.source}'","curl -s https://mcp.noblesight.io/mcp -H \"Authorization: Bearer $KEY\" -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' -d '{\"jsonrpc\":\"2.0\",\"id\":3,\"method\":\"tools/call\",\"params\":{\"name\":\"screen\",\"arguments\":{\"name\":\"Vladimir Putin\",\"lists\":[\"uk\"],\"as_of\":\"2026-07-01T00:00:00Z\"}}}' | sed -n 's/^data: //p' | jq -c '.result.structuredContent|{lists:.lists_screened, omitted:[.warnings[]?|select(.code==\"as_of_before_history\")|.lists[]]}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"lists\":[\"uk\"],\"as_of\":\"2026-07-01T00:00:00Z\"}' | jq -c '{lists:.lists_screened, omitted:[.warnings[]?|select(.code==\"as_of_before_history\")|.lists[]]}'"],"expect":"step 1 prints the tool named screen (no tool named screen_name is listed) with params containing address, as_of, ids and lists, city_never false and city_crimea true; step 2 prints j \"UA-43\" and src \"address\"; steps 3 and 4 print the same line, lists [\"uk\"] and omitted [] (restricting to a list with history at the instant omits nothing); sending as_of alone, with no lists, through either surface names the lists whose history begins after the instant","audit":"the MCP tool is a client of the public API, so a screening it submits is persisted and exportable exactly as one submitted over REST, address, lists, as_of and ids included in the stored request","regression":"TestScreenSchemaMatchesREST, TestScreenForwardsControls, TestScreenToolCallRoundtrip, TestClientMirrorsAPIResponseTypes"}},{"date":"2026-10-25","category":"changed","summary":"A Crimean toponym in address or location.city under a declared country other than Russia or Ukraine is a possible embargo nexus, not a definitive one: jurisdiction_nexus.status \"possible\" names the contradicting field and jurisdiction_risk.level is \"possible_embargo_nexus\"","details":"The 2026-10-24 keyword scan treated every whole-word Crimean toponym in address or location.city as a definitive comprehensive-embargo nexus, so \"Kerch Street 5, Sydney\" under country AU, \"c/o Mr Saky, London\" under GB and \"Sudak Dr, Houston\" under US each returned jurisdiction_risk.level comprehensive_embargo and jurisdiction_nexus.status embargoed, with guidance that most transactions are prohibited, and every such screening became an EDD item to clear by hand. A toponym is evidence of a nexus, not proof of one. Now, when the word is found in address or location.city and the declared country (country or location.country) is neither Russia nor Ukraine, jurisdiction_nexus.status is \"possible\" and a new conflicts array names each contradicting field and value, most specific first; jurisdiction_risk is still issued, naming the same jurisdiction, program and authority, with level \"possible_embargo_nexus\" and guidance that asks for the subject's actual location to be confirmed. The hedge is deliberately narrow, because it is the fail-open direction on a sanctions control: it applies only to a single town name contradicted by a declared country written as a two-letter code. The CoinList shape is unchanged: the same word under country RU or UA, or with no declared country at all, is \"embargoed\" at level comprehensive_embargo. So is corroborated evidence under any country code: the peninsula named outright (Crimea, Krym, Крым) or two different Crimean tokens agreeing across the fields, as in {\"country\":\"US\",\"address\":\"Simferopol, Republic of Crimea\"}. A declared country that is not a two-letter code (Россия, AUS) hedges nothing, and a word in location.region is a declaration and is never hedged. Only declared country fields count as conflicts; the rest of a free-text address is not parsed for a city, state or postcode. The advisory never alters matches, scores or verdict. WHAT YOU WILL SEE: {\"country\":\"AU\",\"address\":\"Kerch Street 5, Sydney NSW 2000, Australia\"} now returns jurisdiction_risk.level possible_embargo_nexus with jurisdiction_nexus {evaluated: Kerch, source: address, status: possible, conflicts: [{source: country, evaluated: AU}]}. Integrations that gate a hard EDD workflow on this signal should key on jurisdiction_risk.level or jurisdiction_nexus.status, not on the block's presence. Replay of a screening run before this change reproduces the definitive nexus it was given, so an earlier record never replays as drift.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"country\":\"AU\",\"address\":\"Kerch Street 5, Sydney NSW 2000, Australia\"}' | jq -c '{j:.jurisdiction_risk.jurisdiction, level:.jurisdiction_risk.level, n:.jurisdiction_nexus}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"country\":\"GB\",\"address\":\"c/o Mr Saky, 12 High Street, London\"}' | jq -c '{j:.jurisdiction_risk.jurisdiction, level:.jurisdiction_risk.level, n:.jurisdiction_nexus}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"country\":\"RU\",\"address\":\"ul. Kirova 1, Simferopol\"}' | jq -c '{j:.jurisdiction_risk.jurisdiction, level:.jurisdiction_risk.level, n:.jurisdiction_nexus}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"country\":\"US\",\"address\":\"ul. Lenina 12, Simferopol, Republic of Crimea 295000\"}' | jq -c '{j:.jurisdiction_risk.jurisdiction, level:.jurisdiction_risk.level, n:.jurisdiction_nexus}'"],"expect":"step 1 prints j \"UA-43\", level \"possible_embargo_nexus\" and nexus {evaluated \"Kerch\", source \"address\", status \"possible\", conflicts [{source \"country\", evaluated \"AU\"}]}; step 2 prints the same shape with evaluated \"Saky\" and conflicts [{source \"country\", evaluated \"GB\"}]; step 3 (the CoinList shape) prints j \"UA-43\", level \"comprehensive_embargo\" and nexus {evaluated \"Simferopol\", source \"address\", status \"embargoed\"} with no conflicts; step 4 (corroborated: town plus the word Crimea under a US country code) prints the same definitive shape as step 3","audit":"the hedged advisory and the nexus with its conflicts are persisted on the screening record like every other nexus, so an export shows the analyst was told possible, not embargoed; the hedge has its own knowledge-timeline floor, so POST /v1/export/replay reproduces the definitive nexus an earlier screening was given rather than reporting drift","regression":"TestResolveNexusConflicts, TestNexusConflictsKnowledgeFloor, TestJurisdictionNexusPossible, TestScreenEmbargoKeywords"}},{"date":"2026-10-24","category":"fixed","summary":"An OFAC match made through a name OFAC publishes in the party's own script is attributed to OFAC SDN: source, prohibition_type, summary.distinct_lists and summary.text all name the list","details":"Since engine 1.2.0 a name is matched against the names OFAC publishes in a party's own script (match_sources source ofac_native_script). When that pathway was the strongest one for a match, the match carried no list label: matches[].source and summary.top_match.source were null, prohibition_type was null where the Latin form of the same party says blocking, and the summary counted the match toward no list, so a confirmed match on SDN 35096 could read 'appears on no sanctions lists' or omit OFAC SDN from distinct_lists. Which text appeared depended on whether another OFAC pathway tied the own-script one on the same screening. The own-script pathway is now labeled OFAC SDN like every other OFAC pathway, so the four fields agree with the Latin form. WHAT YOU WILL SEE: {\"name\":\"Путин Владимир Владимирович\",\"lists\":[\"ofac\"]} returns matches[].source \"OFAC SDN\", prohibition_type \"blocking\", summary.distinct_lists [\"OFAC SDN\"] and a summary.text naming OFAC SDN. No match, score or alert changes; engine_version is unchanged.","endpoint":"/v1/screen","verification":{"steps":["for n in 'Путин Владимир Владимирович' 'Владимир Путин' 'سليم عبد الله سليم الزق'; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n,limit:100}')\" | jq -c --arg n \"$n\" '{n:$n, dl:.summary.distinct_lists, src:.summary.top_match.source, prohib:.matches[0].prohibition_type, ofac_in_text:(.summary.text|contains(\"OFAC SDN\"))}'; done"],"expect":"three lines, each with dl containing \"OFAC SDN\", src \"OFAC SDN\", prohib \"blocking\" and ofac_in_text true; the two Cyrillic lines are SDN 35096 and the Arabic line SDN 57717","audit":"the stored screening record and its replay are unchanged; the list label is derived from the recorded match pathway at response time, so an export of an earlier own-script match shows the same attribution","regression":"TestListLabelOwnScriptPathways, TestListLabelEveryOFACSourceLabeled, TestListLabelMatchesLegacy, TestProhibitionType"}},{"date":"2026-10-24","category":"fixed","summary":"An as_of before Noble's list history is refused instead of answered as a clean screening; a requested list whose history begins after as_of is named in an as_of_before_history warning; GET /v1/lists reports history_from per list","details":"Noble's point-in-time history for a list begins the instant Noble first imported it. Until now an as_of earlier than that returned 200 with lists_screened empty, no matches and verdict no_significant_match, with no warning: a lookback into a year Noble held no history for read as a completed clean screening. Now an as_of earlier than the history of every requested list is refused with 400 invalid_parameter on as_of, and the message names the earliest instant that can be screened and the list it belongs to. When some requested lists have history at the instant and others do not, the screening runs against the lists that do, the others are omitted from lists_screened as before, and a warning with code as_of_before_history names them, with each list's history start, in lists and in the message. GET /v1/lists gains history_from on every list: the RFC 3339 instant its as_of history begins. For the comprehensive embargo in force on a date before the floor, GET /v1/jurisdictions/{code}?as_of= answers from the legal timeline as before; for decisions Noble itself recorded, POST /v1/export/replay reproduces them. WHAT YOU WILL SEE: {\"name\":\"Ramzan Kadyrov\",\"as_of\":\"2026-01-01\"} now returns 400 {code: invalid_parameter, param: as_of} instead of 200 with an empty lists_screened. Live screenings and as_of instants with history on every list are unchanged.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Ramzan Kadyrov\",\"as_of\":\"2026-01-01\"}' | jq -c '{code,param}'","curl -s \"$BASE/v1/lists\" | jq -c '[.lists[]|select(.history_from==null or .history_from==\"\")]|length'","F=$(curl -s \"$BASE/v1/lists\" | jq -r '[.lists[].history_from]|min'); curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"name\\\":\\\"Ramzan Kadyrov\\\",\\\"as_of\\\":\\\"$F\\\"}\" | jq -c '{lists:(.lists_screened|length), omitted:[.warnings[]|select(.code==\"as_of_before_history\")|.lists[]]}'"],"expect":"step 1 prints {\"code\":\"invalid_parameter\",\"param\":\"as_of\"}; step 2 prints 0; step 3 prints lists 1 or more and, whenever lists is fewer than the number of lists GET /v1/lists serves, omitted names exactly the lists that were not screened","audit":"a refused as_of leaves no screening record, like any other rejected input; a partial lookback's record carries the warning in its stored response, so an export shows which lists it did not cover","regression":"TestAsOfRefusal, TestAsOfHistoryWarning, TestHistoryFrom, TestHistoryFromFailsClosed, TestScreenAsOfBeforeHistoryRefused, TestListsReportHistoryFrom"}},{"date":"2026-10-24","category":"added","summary":"GET /v1/account/coverage reports which enforcement-mapped controls your integration exercises, measured from your own screening requests and alert activity","details":"Most of Noble's controls fire only when the integration sends the field they need: an exact name-and-DOB match needs date_of_birth, an identifier hit needs ids, the Crimea advisory needs address or location, four-eyes closure needs a registered analyst in X-Actor. Noble holds every tenant's screening requests and alert activity in the audit record, so it can now tell a tenant, from its own data, which of those controls it is using. GET /v1/account/coverage?from=YYYY-MM-DD\u0026to=YYYY-MM-DD (inclusive UTC days; default the last 90, at most 366) returns the window, the number of api and batch screenings, per input (name, date_of_birth, country, location, address, entity_type, identifiers, lists, as_of) the count and share of screenings that carried it, alert-workflow counts (opened, closed, closed by a named analyst vs the bare key, four-eyes closures, open overdue, active team members), and a controls list with one entry per enforcement-mapped control: the OFAC release that penalized its absence, the enforcement-series page that proves it, what it requires, the measured share, a status (active when at least nine screenings in ten carry what the control needs, partial, inactive, not_applicable) and one sentence on what to send. Monitoring re-screens are the server's, not the integration's, and are not counted. Read-only; no schema change. Also the coverage_report MCP tool.","endpoint":"/v1/account/coverage","verification":{"steps":["curl -s \"$BASE/v1/account/coverage\" -H \"X-API-Key: $KEY\" | jq -c '{screenings, dob:.inputs.date_of_birth, controls:[.controls[]|{id,status}]}'","A=$(curl -s \"$BASE/v1/account/coverage\" -H \"X-API-Key: $KEY\" | jq .inputs.date_of_birth.count); curl -s -o /dev/null -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Coverage Probe\",\"date_of_birth\":\"1970-01-01\"}'; Z=$(curl -s \"$BASE/v1/account/coverage\" -H \"X-API-Key: $KEY\" | jq .inputs.date_of_birth.count); echo \"$A -\u003e $Z\"","curl -s \"$BASE/v1/account/coverage?from=2026-13-01\" -H \"X-API-Key: $KEY\" | jq -c '{code,param}'"],"expect":"step 1 prints screenings \u003e 0, date_of_birth as {count, share} and seven controls each with a status in active/partial/inactive/not_applicable; step 2 prints N -\u003e N+1; step 3 prints {\"code\":\"invalid_parameter\",\"param\":\"from\"}","audit":"read-only over the tenant's own screening_result, monitoring_alert, case_activity and team rows; nothing is written","regression":"TestParseCoverageWindow, TestCoverageStatus, TestCoverageControls, TestAccountCoverage, TestGetCoverageDecodesReport"}},{"date":"2026-10-24","category":"added","summary":"A Crimean city or the word Crimea in address or location.city raises the embargo advisory whatever the country field says, and jurisdiction_nexus names the field and the word","details":"OFAC penalized CoinList (2023-12-13), Uphold (2023-03-31) and Poloniex (2023-05-01) for the same screening gap: a customer gave a non-embargoed country in the country field and a Crimean city, or the word Crimea, in another field, and the screening tool read only the country field. Until now Noble did the same by design: address was recorded and never parsed, and location.city never raised an advisory. Now, when no declared field carries an embargo, location.region, location.city and address are scanned for a whole word naming the Crimean peninsula or a town on it — Crimea, Krym, Simferopol, Sevastopol, Yalta, Kerch, Feodosia, Yevpatoria, Alushta, Dzhankoy, Bakhchysarai, Sudak, Saky, Armyansk, Krasnoperekopsk, with their common Latin transliterations and Russian and Ukrainian Cyrillic forms. A hit raises jurisdiction_risk for UA-43 (Crimea) or UA-40 (Sevastopol) exactly as a declared region would, outranks an unembargoed country and an explicit exclusion, and never outranks a declared embargoed field. jurisdiction_nexus reports source \"address\" or \"location.city\" (or \"location.region\" for a phrase the alias table could not resolve whole, such as \"Crimea Republic\") and evaluated as the single word that matched. Whole words only: \"Krymskaya ulitsa\" in Moscow does not fire; the adjective \"Crimean\" is not on the table; the so-called DNR and LNR stay region-name-only. Advisory, as before: no match score changes and nothing is blocked. WHAT YOU WILL SEE: {\"country\":\"RU\",\"address\":\"Simferopol, Crimea\"} now returns jurisdiction_risk.jurisdiction UA-43 with jurisdiction_nexus {source: address, evaluated: Simferopol, status: embargoed}. Replay of a screening run before this change reproduces it without the keyword scan, so an earlier record never replays as drift.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Ivan Petrov\",\"country\":\"RU\",\"address\":\"ul. Lenina 12, Simferopol, Crimea\"}' | jq -c '{j:.jurisdiction_risk.jurisdiction, n:.jurisdiction_nexus}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Ivan Petrov\",\"location\":{\"country\":\"UA\",\"city\":\"Yalta\"}}' | jq -c '{j:.jurisdiction_risk.jurisdiction, n:.jurisdiction_nexus}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Ivan Petrov\",\"country\":\"RU\",\"address\":\"Krymskaya ulitsa 4, Moscow\"}' | jq -c '{j:.jurisdiction_risk.jurisdiction, n:.jurisdiction_nexus}'"],"expect":"step 1 prints j \"UA-43\" with nexus {evaluated \"Simferopol\", source \"address\", status \"embargoed\"}; step 2 prints j \"UA-43\" with source \"location.city\"; step 3 prints j null with nexus {evaluated \"RU\", source \"country\", status \"no_determination\"}","audit":"the advisory and the evaluated nexus are persisted on the screening record as before; the nexus now says which word in which field raised it, so an examiner can see the address, not the country, was decisive","regression":"TestEmbargoKeyword, TestEmbargoKeywordsNameEmbargoedEntries, TestResolveNexusKeywordPrecedence, TestEmbargoKeywordsOnBothTimelines, TestNexusTokenPrecedence, TestJurisdictionNexusStatuses, TestScreenEmbargoKeywords"}},{"date":"2026-10-21","category":"fixed","summary":"An identifier-only screening raises its own alert: two different listed vessels screened by IMO number within five minutes are two alerts, and the alert's screened_name says which identifier was screened","details":"An alert raised by POST /v1/screen deduplicates against an open alert for the same subject screened by the same account within five minutes, so a double-submit returns the existing alert id. Until now the subject was the name field alone; a screening that sent only identifiers (ids or id_number) had an empty subject, so every identifier-only screening in the window folded into one alert: three different listed vessels screened by IMO number returned one alert_id, that alert's screened_name was empty, and the second and third true hits raised no alert and no webhook event. The subject of an identifier-only screening is now the identifiers as sent (\"IMO 9333591\"; several joined with \", \"), which is also the alert's screened_name. The screening record and the screening response are unchanged: screened_name there is still the name as sent, empty for an identifier-only screening, and replay reads the stored request. WHAT YOU WILL SEE: distinct alert_id values for distinct identifiers; the same identifier screened again within five minutes returns the same alert_id; GET /v1/alerts/{id} shows the identifier in screened_name. Alerts created before this change keep their empty screened_name.","endpoint":"/v1/screen","verification":{"steps":["A=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"ids\":[{\"type\":\"IMO\",\"number\":\"9333591\"}]}' | jq -r .alert_id)","B=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"ids\":[{\"type\":\"IMO\",\"number\":\"8511823\"}]}' | jq -r .alert_id)","C=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"ids\":[{\"type\":\"IMO\",\"number\":\"9333591\"}]}' | jq -r .alert_id); echo \"$A $B $C\"","curl -s \"$BASE/v1/alerts/$A\" -H \"X-API-Key: $KEY\" | jq -r .screened_name"],"expect":"A and B are different alert ids, C equals A, and step 4 prints IMO 9333591","audit":"each screening keeps its own record under its trace id; the alert row carries the identifiers screened in screened_name, so an alert for an identifier-only screening names its subject","regression":"TestAlertSubject, TestIdentifierOnlyScreeningsAlertSeparately"}},{"date":"2026-10-20","category":"fixed","summary":"A no-break space or any other Unicode space between the words of a name no longer lowers its score: Kasem Soleimani typed with U+00A0 scores 100 and alerts, as the ASCII spelling does; engine_version 2.4.0","details":"Word, PDF copy-paste, French typography (U+202F) and CJK input methods (U+3000) put a space other than U+0020 between words, and so do payment fields copied from them. Until now the phonetic (Soundex) part of the score read only U+0020 and the tab as word boundaries, so a name typed with a no-break space was coded as one word, lost the phonetic agreement with the listed surname, and could fall below the alert threshold: Kasem Soleimani scored 68.4 instead of 100 and raised no alert, Aleksey Belan 75 instead of 100, and 7 of 29 multi-word true hits in a customer corpus lost their alert. Every Unicode space (no-break, narrow no-break, figure, thin, ideographic, and the rest of the White_Space set) is now a word boundary on both sides of the comparison, so a listed name scores identically however its spaces were typed. WHAT YOU WILL SEE: engine_version \"2.4.0\"; a multi-word name sent with any Unicode space between its words scores exactly as it does with an ordinary space. Names typed with ordinary spaces score as before. This is a change to what the engine finds, so monitored portfolios are re-screened.","endpoint":"/v1/screen","verification":{"steps":["for sp in 0020 00a0 202f 3000; do n=$(python3 -c \"print('Kasem'+chr(0x$sp)+'Soleimani',end='')\"); curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n,lists:[\"ofac\"]}')\" | jq -r --arg sp $sp '\"U+\\($sp) \\(.engine_version) \\(.matches[0].uid) \\(.matches[0].score) alert=\\(.alert_id != null)\"'; done"],"expect":"four lines, each printing engine_version 2.4.0, uid 10481, score 100 and alert=true","audit":"the screening record stores the name exactly as submitted, including the space characters, and the engine version that scored it, so the decision is reproducible on replay","regression":"TestSoundexForm, TestQuerySoundexSetsUnicodeSpaces, TestNameFormsCorpus (channels c20–c23)"}},{"date":"2026-10-20","category":"fixed","summary":"UK ships designated as IMO n (\"NAME\") are found by their name: ARCTIC EXPRESS scores 100 on UK-RUS3702 and alerts","details":"The UK Sanctions List publishes some designated ships (42 in the 7 October 2026 file, all Russia-regime tankers) with a primary name of the form IMO 9333591 (\"ARCTIC EXPRESS\"): the IMO number first, the vessel's name in a parenthetical. Noble indexed that whole string as the name, so screening the vessel by its own name scored 36.8 to 60 against its own UK record and none of the 42 reached the alert threshold; for 16 of them the UK is the only list naming the ship, and a nine-list screen of the exact vessel name returned no alert. The importer now reads that form as the vessel name ARCTIC EXPRESS with IMO number 9333591, so the name is indexed as every other ship's is and the IMO number is screenable as an identifier (see the companion entry). Ships already imported are rewritten on the next UK import after this release, which is forced; each renamed ship is recorded as modified in the UK change history and re-screened against monitored portfolios. Normally named UK ships (FELIKS, MARS 6) are unchanged.","endpoint":"/v1/screen","verification":{"steps":["for n in 'ARCTIC EXPRESS' 'GALLE ENERGY' 'KONSTANTIN POSIET' 'FELIKS'; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n,lists:[\"uk\"],limit:100}')\" | jq -r --arg n \"$n\" '\"\\($n): \\(.matches[0].uid) \\(.matches[0].score) \\(.matches[0].last_name) alert=\\(.alert_id != null)\"'; done"],"expect":"ARCTIC EXPRESS → UK-RUS3702 100, GALLE ENERGY → UK-RUS3751 100, KONSTANTIN POSIET → UK-RUS3737 100, FELIKS → UK-RUS2765 100, each with last_name equal to the vessel name (no IMO prefix) and alert=true","audit":"the rename of each ship is a modified entry in Noble's UK list change history carrying the composite name before and the vessel name after, available from support@noblesight.io on request; screenings recorded before the import keep the names they were scored against, and replay reports the engine change","regression":"TestSplitVesselName, TestParseXMLVesselNameForm, TestParseXMLLiveFixtureVesselNames"}},{"date":"2026-10-20","category":"fixed","summary":"UK and Canada ships are screenable by IMO number: {\"ids\":[{\"type\":\"IMO\",\"number\":\"9333591\"}]} returns UK-RUS3702 at 100 and alerts; a Canada-only ship's IMO returns its Canada record","details":"Identifier screening read each list's identity-document table, and the UK and Canada publish a ship's IMO number elsewhere: the UK in the ship's details block, Canada as a column of the designation itself. An IMO number designated only by one of those lists therefore returned match_count 0 across all nine lists, indistinguishable from a clean vessel, while OFAC IMO numbers matched at 100. Both are now read: a UK or Canada IMO number returns the ship at 100 with match source uk_id or canada_id, bare or with the IMO prefix, with or without type IMO (OFAC's category label Vessel Registration Identification is accepted as a synonym). A typed request for any other category still never reaches an IMO row. The default list set for identifier screening now includes canada; lists restricts it as before.","endpoint":"/v1/screen","verification":{"steps":["for imo in 9333591 9459242 9253909 8511823; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg i \"$imo\" '{ids:[{type:\"IMO\",number:$i}],limit:100}')\" | jq -r --arg i \"$imo\" '\"\\($i): \\(.match_count) \\(.matches[0].uid) \\(.matches[0].score) \\(.matches[0].match_sources[0].source) alert=\\(.alert_id != null)\"'; done"],"expect":"9333591 → UK-RUS3702, 9459242 → UK-RUS2765, 9253909 → UK-RUS2719, each at 100 with source uk_id and alert=true; 8511823 → 19585 at 100 with source ofac_id and alert=true (the OFAC path is unchanged)","audit":"an identifier match is recorded with the identifier submitted and the list record it hit, under the list versions cited on the screening","regression":"TestVesselIMOScreenable, TestIDScreenersMatchLegacy"}},{"date":"2026-10-20","category":"fixed","summary":"Canada ships are typed vessel: MARJORIE with entity_type vessel scores 100 on CA-Russia-1.1-659 and alerts instead of 50 with no alert","details":"Canada's SEMA list publishes no subject type. Noble typed every Canada record with a name in the entity-or-ship column as entity, so a caller who correctly sent entity_type vessel for a Schedule 1.1 ship was penalised for the mismatch (−50) and an exact-name true hit fell from 100 to 50 with no alert, while a caller who sent nothing, or the wrong type, got 100 and an alert. A Canada record carrying a ship IMO number is now typed vessel (every Schedule 1.1 vessel carries one; no other row does), so sdn_type reads vessel, entity_type vessel matches it, and entity_type entity or individual is now the mismatch. Other Canada records are typed as before. WHAT YOU WILL SEE: sdn_type \"vessel\" on Canada ship matches; a request with entity_type vessel scores a Canada ship as a request without entity_type does.","endpoint":"/v1/screen","verification":{"steps":["for et in vessel entity none; do b='{\"name\":\"MARJORIE\",\"lists\":[\"canada\"],\"limit\":100'; [ $et != none ] \u0026\u0026 b=\"$b,\\\"entity_type\\\":\\\"$et\\\"\"; b=\"$b}\"; curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$b\" | jq -r --arg et $et '\"\\($et): \\(.matches[0].uid) \\(.matches[0].score) \\(.matches[0].sdn_type) alert=\\(.alert_id != null)\"'; done"],"expect":"vessel: CA-Russia-1.1-659 100 vessel alert=true; none: the same at 100; entity: 50 with entity_type_mismatch and no alert","audit":"the screening record stores the entity_type submitted and each match's dismissal signals, so the penalty applied, or not, is reproducible","regression":"TestCanadaShipTypedVessel"}},{"date":"2026-10-20","category":"added","summary":"The account's entitlement record is readable: GET /v1/account/features/history lists every grant and withdrawal of the portfolio-monitoring and Confidential Portfolio add-ons","details":"A compliance officer asked when portfolio monitoring was on for an account, who turned it on and why, can now answer from the API rather than from a support request. Each row is one change to portfolio_monitoring or confidential_portfolio on any of the account's keys, as recorded: feature, value before and after, who made it (operator:noblesight for a Noble operator acting on a documented reason, stripe:checkout for the customer's own purchase, customer:\u003cclient_id\u003e for the customer's own action), the reason given, the trace id of the request, the key prefix, and the timestamp. Rows come from an append-only record with a single writer and are under the same daily retention verification as the screening record. The list is cursor-paginated oldest first and filters by feature and by day (from/to). The current state remains features on GET /v1/account. Available on every tier. The same record is available to agents through the MCP tool list_feature_history.","endpoint":"/v1/account/features/history","verification":{"steps":["curl -s \"$BASE/v1/account/features/history\" -H \"X-API-Key: $KEY\" | jq -c '.data[] | {id, feature, enabled_before, enabled_after, changed_by, reason, changed_at}'","curl -s \"$BASE/v1/account/features/history?limit=1\" -H \"X-API-Key: $KEY\" | jq -c '{n: (.data|length), has_more}'","curl -s -o /dev/null -w '%{http_code}\\n' \"$BASE/v1/account/features/history?feature=bogus\" -H \"X-API-Key: $KEY\"","curl -s \"$BASE/v1/account\" -H \"X-API-Key: $KEY\" | jq -c '.features | {portfolio_monitoring, confidential_portfolio}'"],"expect":"step 1 prints one line per change in ascending id order, each with changed_by equal to operator:noblesight, stripe:checkout or customer:\u003cyour client_id\u003e and a non-empty reason, or nothing if the add-ons have never changed; step 2 prints n of at most 1 and has_more true exactly when more than one change exists; step 3 prints 400; the last enabled_after for each feature in step 1 equals that feature's value in step 4","audit":"the rows are the record: keep the step 1 output with your own compliance files; each id is the audit id Noble support cites, and each trace_id joins to the request that made the change","regression":"TestFeatureHistoryEndToEnd, TestFeatureHistoryClientIsolation"}},{"date":"2026-10-19","category":"added","summary":"Portfolio monitoring and Confidential Portfolio add-ons are granted and withdrawn through one audited path","details":"A trial or comp of the portfolio-monitoring add-on no longer depends on a manual database change by Noble. A grant or withdrawal takes effect on the customer's next request: GET /v1/account reports it under features.portfolio_monitoring and features.confidential_portfolio, and POST /v1/portfolio answers 201 instead of 403 tier_insufficient. Every change to either add-on, whether made by a Noble operator, by Stripe checkout, or by the customer destroying their Confidential Portfolio key, is written to an append-only record naming who made it, the reason given, and the value before and after. Withdrawing Confidential Portfolio while the customer still holds an active encryption key is refused, so encrypted data is never left under a key the customer can no longer destroy. The response shapes of GET /v1/account and POST /v1/portfolio did not change.","endpoint":"/v1/account","verification":{"steps":["curl -s \"$BASE/v1/account\" -H \"X-API-Key: $KEY\" | jq -c '.features | {portfolio_monitoring, confidential_portfolio}'","curl -s -o /dev/null -w '%{http_code}\\n' -X POST \"$BASE/v1/portfolio\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"Add-on Probe\"}'"],"expect":"step 1 prints both flags as booleans; step 2 prints 201 when portfolio_monitoring is true and 403 when it is false, and the two agree on the first request after any grant or withdrawal","audit":"every grant or withdrawal is an append-only record carrying the actor, the reason, the previous and new value, and the request's trace id; ask support@noblesight.io for your account's entitlement history","regression":"TestAdminTenantFeaturesGrantRevoke, TestAdminTenantFeaturesRefusesRevokeWithActiveKey"}},{"date":"2026-10-18","category":"changed","summary":"Public documentation corrected: nine live lists, invite-only keys, what a non-Latin name is compared against, and what ownership \"unknown\" means","details":"llms.txt, the marketing page and the docs page now state the facts the API already reported. GET /v1/lists serves nine lists (OFAC SDN and Consolidated, UK, EU, UN, France, Belgium, Netherlands, Canada, Australia); earlier copy said eight or ten. POST /v1/keys requires an invite code while Noble is in beta, and the copy that promised a key with no invite is gone; requests go to sales@noblesight.io. A name written in Cyrillic, Arabic, Hangul or another script is compared against the names the lists publish in that script, including OFAC's, and the script_not_screened warning says which part was not compared against Latin-only parties; the earlier text said OFAC publishes Latin names only. An ownership status of unknown means the name resolved to no party in the ownership graph; it is not a clean determination. Uptime monitors should use GET on /status. No screening behaviour changed.","endpoint":"/v1/keys","verification":{"steps":["curl -s \"$BASE/v1/lists\" | jq '.lists | length'","curl -s \"$BASE/llms.txt\" | grep -c 'nine live global sanctions lists'","curl -s -X POST \"$BASE/v1/keys\" -H 'Content-Type: application/json' -d '{\"client_id\":\"doc-check\",\"accept_terms\":true}' | jq -r .code","curl -s -o /dev/null -w '%{http_code}\\n' -I \"$BASE/status\"; curl -s -o /dev/null -w '%{http_code}\\n' \"$BASE/status\""],"expect":"step 1 prints 9; step 2 prints a count of at least 1; step 3 prints invite_code_required; step 4 prints 405 for HEAD and 200 for GET, which is why the copy says to monitor with GET","audit":"documentation only; no screening record, score or alert changes"}},{"date":"2026-10-18","category":"changed","summary":"A name over 500 characters is refused with code name_too_long on every endpoint","details":"A name longer than 500 characters is refused with 400 and code name_too_long, param naming the field, on POST /v1/screen, POST /v1/batch, POST /v1/watchlist and POST /v1/portfolio. Until now the same refusal carried name_required on /v1/screen and /v1/batch, which reads as a missing name, and invalid_parameter on /v1/watchlist and /v1/portfolio, so a client had to match three codes for one rule; name_too_long was published but never returned. The limit itself, counted in characters, is unchanged. A client that matched name_required or invalid_parameter for an over-long name should match name_too_long.","endpoint":"/v1/screen","verification":{"steps":["N=$(python3 -c \"print('A'*501)\")","for E in screen watchlist portfolio; do F=full_name; [ $E = screen ] \u0026\u0026 F=name; curl -s -X POST \"$BASE/v1/$E\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"$F\\\":\\\"$N\\\"}\" | jq -r --arg e $E '$e + \" \" + .code'; done","curl -s -X POST \"$BASE/v1/batch\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"names\\\":[{\\\"full_name\\\":\\\"$N\\\"}]}\" | jq -r '\"batch \" + .code'"],"expect":"every line prints name_too_long: \"screen name_too_long\", \"batch name_too_long\", \"watchlist name_too_long\", and \"portfolio name_too_long\" (portfolio needs the monitoring add-on; without it that line prints tier_insufficient, which is checked before the body)","audit":"a refused request is not screened and writes no screening record, as before","regression":"TestNameTooLongCodeAgrees, TestCheckScreeningName"}},{"date":"2026-10-18","category":"fixed","summary":"GET /v1/batch/status reports a batch's progress while its names are being screened","details":"GET /v1/batch/status returned 500 internal_error (\"batch results lookup failed\") for as long as any name in the job was still waiting to be screened, so a job could only be read once every name had been screened, and processed_names never showed progress. It now returns 200 at every stage: the job's status, total_names and processed_names, and every item, with status \"pending\" on each name not yet screened. Screening, results and alerts are unchanged; only reading a job mid-run was affected.","endpoint":"/v1/batch/status","verification":{"steps":["ID=$(python3 -c \"import json;print(json.dumps({'names':[{'full_name':'Zzyxx Probe %d' % i} for i in range(300)]}))\" | curl -s -X POST \"$BASE/v1/batch\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d @- | jq -r .id)","curl -s -o /dev/null -w '%{http_code}\\n' \"$BASE/v1/batch/status?id=$ID\" -H \"X-API-Key: $KEY\"; curl -s \"$BASE/v1/batch/status?id=$ID\" -H \"X-API-Key: $KEY\" | jq -c '{status, total_names, processed_names, pending: ([.items[] | select(.status == \"pending\")] | length)}'"],"expect":"the status code is 200, and the body shows total_names 300 with status pending or processing while names remain (pending \u003e 0 on an early poll); polled again until status is completed, processed_names is 300 and pending is 0 (Standard tier or higher, which batch requires)","audit":"reading a job changes nothing: items, their results and the job record are written only by the screening worker","regression":"TestBatchStatusWhileItemsPending"}},{"date":"2026-10-18","category":"fixed","summary":"The hosted MCP screening tool returns every field /v1/screen does, including warnings and engine_version","details":"The screen_name tool on mcp.noblesight.io now returns the same result /v1/screen does. Until now it dropped fields /v1/screen added after its last release: warnings (such as script_not_screened, which says a name's script was not compared against lists that publish only Latin names), engine_version and match_engine on screenings, and scan_pin and the linked alerts on stored screening records. So an agent could read a name as screened and clear when part of it had not been compared, with nothing in the result saying so. No match, score or alert changes; the tool now reports what the API already decided.","endpoint":"/v1/screen","verification":{"steps":["In any MCP client connected to https://mcp.noblesight.io/mcp with your API key, call the screen_name tool with {\"name\":\"Іван Тестовський\"}","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Іван Тестовський\"}' | jq -c '{warnings: [.warnings[]?.code], engine_version, match_engine}'"],"expect":"the tool result's warnings include script_not_screened and it carries engine_version and match_engine, the same three values step 2 prints from /v1/screen","audit":"the MCP tool is a client of the public API, so what it returns is what the screening record holds","regression":"TestClientMirrorsAPIResponseTypes"}},{"date":"2026-10-18","category":"fixed","summary":"match_sources report trigram_score to one decimal, like every other score","details":"Each entry in a match's match_sources now reports trigram_score to one decimal place, as score and soundex_boost already were, and trigram_score plus soundex_boost equals the pathway's contribution to one decimal. Until now trigram_score carried the full binary fraction, so evidence read 55.00000000000001 or 42.30769230769231. No match, score, ranking or alert changes; only how the per-source figure is printed.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimer Puttin\"}' | jq -c '[.matches[].match_sources[] | .trigram_score, (.soundex_boost // 0)] | map(tostring | select(test(\"\\\\.[0-9]{2,}\"))) | length'"],"expect":"prints 0: every trigram_score and soundex_boost has at most one decimal place","audit":"stored screening records keep the figures they were written with; replay compares matches and scores, not match_sources","regression":"TestMatchSourceScoresRounded"}},{"date":"2026-10-18","category":"fixed","summary":"as_of screenings and replays read OFAC aliases and native-script names as they stood, and say so whenever OFAC is involved","details":"An as_of screening, and POST /v1/export/replay of a stored screening, read OFAC as it stood at the screening's own point in time. Until now two kinds of OFAC name were read as they stand today: an alias OFAC published later was screened at the earlier point, and a name in a party's own script that OFAC has since removed was not. Both are now read as they stood: a later alias is not screened, and a removed native-script name still is, so a replay of a record made before either change reproduces the record. The ofac_child_detail_current_state caveat on a replay now appears whenever either the stored record or the replay has an OFAC match, not only when the stored record did, so an OFAC match found only by the replay is never reported without it. An as_of screening that screens OFAC now carries a warning with the same code, ofac_child_detail_current_state, stating what OFAC data an as_of result reads at today's state. No live POST /v1/screen match, score or alert changes, and engine_version is unchanged.","endpoint":"/v1/export/replay","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"lists\":[\"ofac\"],\"as_of\":\"2026-09-15\"}' | jq -c '[.warnings[]?.code]'","curl -s \"$BASE/v1/export/bulk?from=2026-09-01\u0026to=2026-09-30\u0026channel=api\u0026limit=100\" -H \"X-API-Key: $KEY\" | jq -r '.data[] | select(.outcome == \"completed\") | .trace_id' | while read t; do curl -s -X POST \"$BASE/v1/export/replay\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"trace_id\\\":\\\"$t\\\"}\" | jq -c '{ofac_any: ([(.replayed.matches[]?.uid), (.stored.matches[]?.uid)] | map(select(test(\"^[0-9]+$\"))) | length \u003e 0), caveat: ((.caveats // []) | index(\"ofac_child_detail_current_state\") != null)}'; done | jq -s '[.[] | select(.ofac_any and (.caveat | not))]'"],"expect":"the first command's list includes \"ofac_child_detail_current_state\"; the second prints [] — every replay with an OFAC match on either side carries the ofac_child_detail_current_state caveat (Standard tier or higher, which replay requires)","audit":"names OFAC removes from its own-script publication are kept on record with the moment they were removed, and each alias records the OFAC list version that first published it, so a historical screening and its replay read the same OFAC names the record did","regression":"TestReplayCutBeforeAlias, TestReplayCutBeforeNativeRemoved, TestCutBeforeAliasInsertExcludesIt, TestCutBeforeNativeRemovalKeepsIt, TestCompareReplayOFACCaveatOnEitherSide, TestOFACChildCutoffs, TestSnapshotAtLiveVectorMatchesLive"}},{"date":"2026-10-18","category":"fixed","summary":"A name whose accented letters were garbled by a wrong-encoding read, such as JoaquÃ­n GuzmÃ¡n from a CSV opened with the wrong character set, is now screened as the name it was meant to be; engine_version 2.3.0","details":"When text written in UTF-8 is read as Latin-1 or Windows-1252 somewhere before it reaches Noble, each accented letter turns into two or three other characters: Joaquín becomes JoaquÃ­n, O’Brien becomes Oâ€™Brien. Until now the engine screened the garbled text as sent, so a listed name damaged this way scored well below its listed form and could fail to alert. Now, when a name reads as exactly this kind of damage, the engine also screens the repaired name, and each listed party keeps its best score across the two. The checks that run after matching (the one that discounts a match on only one of a person's names, and the near-miss signal) read the repaired name too, so the repaired name gets the same final score and dismissal signals as the correctly typed one. The name as you sent it is always screened too, so a repair can add a match but never remove one, and a name typed correctly with accents (Müller, Joaquín) is never changed. Text garbled twice is repaired too. WHAT YOU WILL SEE: engine_version \"2.3.0\"; \"JoaquÃ­n GuzmÃ¡n Loera\" scores 100 against the same record as \"Joaquín Guzmán Loera\" and alerts; screened_name still shows the name exactly as you sent it. Names without this damage score as before. This is a change to what the engine finds, so monitored portfolios are re-screened.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"JoaquÃ­n GuzmÃ¡n Loera\",\"lists\":[\"ofac\"]}' | jq -c '{engine_version, uid:.matches[0].uid, score:.matches[0].score, alerted:(.alert_id != null)}'"],"expect":"prints engine_version \"2.3.0\", uid \"6861\", score 100 and alerted true","audit":"the screening record stores the name exactly as submitted and the engine version that scored it, so the repair is reproducible on replay","regression":"TestRepairMojibake, TestQueryFormsMojibake, TestControlsJudgeRepairedName, TestNameFormsCorpus, TestFalsePositiveBenchmark, TestSeededMissBenchmark, TestRegionalRecallBenchmark"}},{"date":"2026-10-18","category":"changed","summary":"Internal watchlist entries are compared in the same normalized form as the sanctions lists, so an entry added with accents matches its plain spelling and the reverse; engine_version 2.2.0","details":"The normalization the engine applies to every sanctions-list name and every screened name — accents, apostrophes, punctuation and lookalike letters do not change a score — now applies to your internal watchlist too. An entry added as \"José María Álvarez\" is matched by \"Jose Maria Alvarez\" and \"JOSE MARIA ALVAREZ\" at the same score as its own spelling, and an entry added without accents is matched by an accented query. Entries you added before this release are normalized automatically and match the same way from the moment it is live. WHAT YOU WILL SEE: engine_version \"2.2.0\"; internal watchlist matches (source \"Internal Watchlist\") for spellings of an entry that differ only by accents, apostrophes or letter case, at the score of an exact spelling; and a watchlist match carried partly by phonetic agreement now reports soundex_boost in its match_sources entry, as the sanctions pathways do, so trigram_score plus soundex_boost accounts for its score. Nothing changes for entries and queries already typed identically.","endpoint":"/v1/watchlist","verification":{"steps":["ID=$(curl -s -X POST \"$BASE/v1/watchlist\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"José María Álvarez\",\"list_name\":\"SAR Subjects\"}' | jq -r .id)","for n in 'José María Álvarez' 'Jose Maria Alvarez' 'JOSE MARIA ALVAREZ'; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n}')\" | jq -c --arg id \"WL-$ID\" '[.engine_version, (.matches[] | select(.uid == $id) | .score)]'; done"],"expect":"each of the three lines prints [\"2.2.0\",100]","audit":"every screening record stores the engine version that scored it; watchlist matches carry the entry's id and the text they were compared on","regression":"TestWatchlistFoldedBothSides, TestWatchlistRetrievalSuperset, TestRefoldWatchlist, TestRefoldWatchlistTerminates, TestWatchlistReportsSoundexBoost"}},{"date":"2026-10-18","category":"changed","summary":"Names are compared in one normalized form, so how a name was typed — accents, curly apostrophes, passport machine-readable fillers, lookalike letters from another script — no longer changes its score; engine_version 2.1.0","details":"Before comparing a name with the sanctions lists, the engine now normalizes both sides the same way: accents removed (Joaquín becomes joaquin, Ališer becomes aliser), ligatures and fullwidth letters expanded, apostrophe and quote variants removed, dashes, punctuation and the \u003c filler of a passport's machine-readable zone read as spaces, invisible characters removed, and a Cyrillic or Greek letter that looks like a Latin one, inside a word written in Latin, read as that Latin letter. A name written entirely in another script stays in that script. Phonetic matching reads the normalized word too, so a listed surname typed with an accent or a curly apostrophe earns the same phonetic credit as its plain spelling, and a hyphen typed as a non-breaking hyphen or a dash (as word processors and PDFs often do) counts as a hyphen, so \"Kim Chong‐un\" matches as \"Kim Chong-un\" does, and the check that discounts a match on only one of a person's names reads both names the same way, so an accent can no longer make a full-name match look like a partial one. A name containing an invisible character between two words is also screened with it read as a space. WHAT YOU WILL SEE: engine_version \"2.1.0\"; the accented and plain spellings of a listed name score the same — \"Joaquín Guzmán Loera\" and \"Joaquin Guzman Loera\" both score 100 against the same record and alert; more matches where a name was typed with or without accents; matched_text still shows the name as the list publishes it. Scores for names already typed in plain ASCII do not change. This is a change to what the engine finds, so monitored portfolios are re-screened.","endpoint":"/v1/screen","verification":{"steps":["for n in 'Joaquín Guzmán Loera' 'Joaquin Guzman Loera' 'Ališer Usmanov' 'Aliser Usmanov'; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n,limit:3}')\" | jq -c --arg n \"$n\" '{n:$n, engine_version, top:[.matches[0:2][]|{uid,score}]}'; done"],"expect":"every line prints engine_version \"2.1.0\"; both Guzmán lines print uid \"6861\" first at score 100; the two Usmanov lines list the same two records at the same scores","audit":"every screening record stores the engine version that scored it; replaying a record scored before 2.1.0 carries the engine_version_changed caveat rather than reporting the change as data drift","regression":"TestSoundexForm, TestSoundexDashesAreHyphens, TestRecordSurnameCodesFoldFirst, TestBoostCodes, TestNameTokenSignalReadsFoldedNames, TestQueryFormsInvisibleCharacters, TestIndexFoldsBothSides, TestNameFormsCorpus, TestFalsePositiveBenchmark, TestRegionalRecallBenchmark, TestSeededMissBenchmark"}},{"date":"2026-10-18","category":"changed","summary":"Every name screening now runs on Noble's in-memory matching engine: engine_version 2.0.0, match_engine \"index\"; names are limited to 500 characters rather than bytes, and text that is not valid UTF-8 is refused","details":"Until now, name screening compared names inside the database. From this release, one engine decides every name match: live screening, POST /v1/batch, deep_screen's variation screens, as_of screenings, POST /v1/export/replay, and the internal watchlist. It compares names in memory against a snapshot of the sanctions lists, and every screening states which snapshot in list_versions and scan_pin. The engine's rules are a published specification rather than a database extension, and the screening corpus, recall and false-positive benchmarks run through it on every change. WHAT YOU WILL SEE: engine_version \"2.0.0\" and match_engine \"index\" on every screening. Names written in Latin script score as before. Names in other scripts can score slightly differently, because letters are now compared as letters rather than as database byte sequences. Each POST /v1/batch item is scored against the lists current when it is screened, and its record states that list version, so a job that runs across a list update moves to the new lists. An as_of screening or a replay of an older record may take a few seconds the first time a list version is requested; as_of instants between the same two list updates share one prepared version, and while a server is preparing another one it may answer 503 coverage_unavailable with a Retry-After header. If a server cannot load the lists, or cannot confirm that the lists it holds are current, name screening (POST /v1/screen, POST /v1/batch, POST /v1/export/replay) refuses with 503 coverage_unavailable and a Retry-After header rather than screening any other way, while alerts, exports and the rest of the API keep serving; GET /status then reports components.screening as \"degraded\" or \"outage\". This is rare and safe to retry after the Retry-After interval. Every part of one screening comes from the one list state its list_versions cite: the name's matches, each deep_screen variation's, identifier matches, the date-of-birth and country comparisons, and the programs, addresses and identifiers on each match. A list update that lands while a request runs is never mixed into it; the next screening sees it. Names may be up to 500 characters long, counted as characters, so a name in Cyrillic or Arabic gets the same room as one in Latin; the same limit, counted the same way, applies to POST /v1/watchlist and POST /v1/portfolio. A request containing bytes that are not UTF-8 text is refused with 400 invalid_parameter naming the field (param name on /v1/screen) instead of being screened as something you did not send; on POST /v1/batch only the item holding them is affected: the job is accepted, that item is reported with status error and error_code invalid_parameter, and every other item is screened. A name holding the U+FFFD replacement character, which is valid UTF-8 and is what a system writes where it already lost a letter, is screened, and the response carries a name_replacement_character warning. Batch names are trimmed of surrounding whitespace, as /v1/screen names already were, and an empty one is refused as required (missing_parameter). deep_screen screens at most 24 variations per request (MaxDeepScreenVariations), each held to the same name rule. Replaying a record scored by the previous engine carries the engine_version_changed caveat, and so does a record stored before engine_version was recorded; its scores are then not compared, but its decision is: a summary.verdict that differs from the record's is reported as a difference and the replay verdict is mismatch, never match. Monitored portfolios are re-screened under the new engine. We also no longer state that a screening completes in under 100 milliseconds: our published target is a 110 ms median measured server-side, and every response's Server-Timing header shows the server's own measurement.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"lists\":[\"ofac\"]}' | jq -c '{engine_version, match_engine, uid:.matches[0].uid, score:.matches[0].score}'","printf '{\"name\":\"Vladimir \\xff Putin\"}' | curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' --data-binary @- | jq -c '{code, param}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Put\\ufffdn\",\"lists\":[\"ofac\"]}' | jq -c '{code, warn:[.warnings[]?.code], uid:.matches[0].uid}'","printf '{\"names\":[{\"full_name\":\"Vladimir Putin\"},{\"full_name\":\"Jos\\xe9 Garc\\xeda\"}]}' | curl -s -w ' %{http_code}\\n' -X POST \"$BASE/v1/batch\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' --data-binary @-","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$(printf 'Ж%.0s' $(seq 1 500))\" '{name:$n}')\" -o /dev/null -w '%{http_code}\\n'","for d in 2025-01-01 2025-01-02 2025-01-03; do curl -s -o /dev/null -w \"%{http_code} %header{retry-after}\\n\" -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"name\\\":\\\"Vladimir Putin\\\",\\\"as_of\\\":\\\"${d}T00:00:00Z\\\"}\" \u0026 done; wait","curl -s \"$BASE/status\" | jq -r .components.screening.status"],"expect":"the first command prints engine_version \"2.0.0\", match_engine \"index\", uid \"35096\" and score 100; the second prints code \"invalid_parameter\" and param \"name\"; the third prints code null, warn [\"name_replacement_character\"] and uid \"35096\"; the fourth prints the job (its id) followed by 202, and GET /v1/batch?id=\u003cid\u003e once the job completes shows the second item with status \"error\" and error_code \"invalid_parameter\" and the first screened; the fifth prints 200 (500 Cyrillic characters, 1,000 bytes, are accepted); the sixth prints one line per date, each 200, or 503 followed by a non-empty Retry-After value (never 502 or 504); the seventh prints \"operational\"","audit":"every screening record stores the engine version and engine that scored it and the list versions it was scored against; a replay rebuilds the lists at those versions and scores them with the same engine","regression":"TestScreeningReportsAndStoresEngineVersion, TestLiveIndexRefreshesOnChange, TestSnapshotAtLiveVectorMatchesLive, TestCutIndexesCache, TestCutIndexesCapConcurrentBuilds, TestCutIndexesShareOneVector, TestLiveIndexStaleWhenFingerprintFails, TestIndexUnavailableScopedToScreening, TestWatchlistRetrievalSuperset, TestCheckScreeningName, TestScreenRefusesInvalidUTF8, TestScreenScreensValidUTF8ReplacementCharacter, TestBatchInvalidUTF8ItemReportedAlone, TestBatchEmptyNameIsRequired, TestWatchlistAcceptsNameCountedInCharacters, TestDeepScreenVariationsCapped, TestDeepScreenVariationsScoredOnTheNamesIndex, TestIdentifierMatchReadAtTheIndexPin, TestReplayOfRecordPreEngineVersion, TestReplayUnderEngineChangeReportsVerdictChange, TestReplayComparePreEngineVersion, TestReplayCompareVerdictChange, TestNameFormsCorpus, TestFalsePositiveBenchmark, TestRegionalRecallBenchmark"}},{"date":"2026-10-18","category":"changed","summary":"Portfolio monitoring compares folded names and scores with the same rules as /v1/screen; engine_version 1.3.0","details":"When a sanctions list changes, Noble compares the changed names against your monitored portfolio. That comparison now uses each customer's name_fold, the folded form shown on GET /v1/portfolio, against the changed name folded the same way, so case, accents, ligatures, apostrophes and lookalike letters no longer separate a customer from a listed party: a customer enrolled as Ališer Usmanov is compared with a listed ALISHER USMANOV on equal terms, and a listed name published with accents reaches a customer enrolled without them. The database now only finds the portfolio rows near each changed name; every decision and score is made by Noble's scorer with the rules /v1/screen applies, so a customer's name scores the same in monitoring as it does on /v1/screen. A listed party is compared by its full name, its surname and its given name, and the closest of the three counts, on its primary name and on every alias, so a customer enrolled as Guzman Loera is found by the listing of Joaquin GUZMAN LOERA. An alias is credited for a surname that sounds like the listed party's own, as on /v1/screen. A customer name that carries a bracketed note, or whose words are separated by invisible characters such as zero-width spaces, is compared in each form /v1/screen screens it in, and the closest form counts. A portfolio row whose folded form has not been written yet, or was written by an earlier version of the fold, is still screened: it is folded when the listing is compared and decided by the same rules. A monitoring alert's match_payload reports SoundexMatch true whenever the surname agreed phonetically, as /v1/screen reports soundex_match, including when the name already matched in full. Names held in a Confidential Portfolio are compared in memory after decryption, exactly as plaintext names are, and no decrypted name is sent back to the database. Because this changes what monitoring finds, engine_version becomes 1.3.0 on every response and record, monitoring evidence records carry match_engine \"scorer\", and monitored portfolios are re-screened: alerts from that re-screen carry change_type engine_change. On /v1/screen, a name whose words are separated by invisible characters is now also screened with those characters read as spaces; every other /v1/screen result is unchanged in this release. Replaying a screening recorded under 1.2.0 reports the engine_version_changed caveat.","endpoint":"/v1/portfolio","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\"}' | jq -c '{engine_version, match_engine}'","curl -s \"$BASE/v1/alerts?limit=20\" -H \"X-API-Key: $KEY\" | jq -c '.data[] | select(.change_type == \"engine_change\") | {id, change_type, screened_name, score}'"],"expect":"the first command prints engine_version \"1.3.0\" and match_engine \"database\" (the live screen is unchanged); after the re-screen, any monitoring alert it raised for your portfolio lists change_type \"engine_change\"","audit":"every monitoring decision made under these rules is stored with engine_version 1.3.0 and match_engine \"scorer\" beside the alert it produced, so a record always names the rules that made it","regression":"TestReverseForwardParity, TestReverseParityPairs, TestReverseRetrievesUnfoldedRows, TestReverseSoundexMatchAtCappedScore, TestReverseAliasBoostUsesDesignationCodes, TestReverseRetrievalSuperset, TestReverseRegionalRecall, TestReverseFalsePositiveRate, TestReverseProbeScore, TestReverseRetrievalNeedsNoSoundexArm, TestReverseScreenEncryptedScoresInMemory, TestReverseScreenScale, TestQueryForms"}},{"date":"2026-10-15","category":"fixed","summary":"Portfolio monitoring now screens the names OFAC and the UK publish in a party's own script, as POST /v1/screen already matches them","details":"POST /v1/screen matches a name against the names a list publishes in the party's own script (match_source ofac_native_script for OFAC, uk_non_latin for the UK). Portfolio monitoring did not always screen your portfolio against those names when a list changed. OFAC publishes them in a separate publication on its own schedule; when it carried a new party's own-script names before the OFAC list added the party, those names were never screened against portfolios. UK own-script names were never screened against portfolios, and a UK change to them alone was not treated as a list change. Now every OFAC list change that adds or updates a party carries the party's current own-script names, every UK list change carries the designation's own-script names, and a UK change to those names alone is a list change that is screened against every monitored portfolio. A name you monitor written in Cyrillic, Arabic or another script raises the same monitoring alert its POST /v1/screen result would. No match, score or response field changes on POST /v1/screen.","endpoint":"/v1/portfolio","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Путин Владимир Владимирович\",\"lists\":[\"uk\",\"ofac\"]}' | jq -c '[.matches[] | {uid, match_source, score}]'","curl -s -X POST \"$BASE/v1/portfolio\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"Путин Владимир Владимирович\"}' | jq -c '{id, full_name}'","curl -s \"$BASE/v1/alerts?source=monitoring\u0026limit=100\" -H \"X-API-Key: $KEY\" | jq -c '[.data[] | select(.screened_name == \"Путин Владимир Владимирович\") | {id, screened_name, score}]'"],"expect":"the first command lists matches with match_source \"ofac_native_script\" or \"uk_non_latin\"; the second returns 201 with an id; after the next OFAC or UK list change that adds or updates a party whose own-script name matches it, the third lists a monitoring alert for the enrolled name whose score is within 0.1 of that party's score in the first command","audit":"the portfolio-monitoring job a list change becomes records every name it screened, own-script names included, so an examiner can see each published name was screened against portfolios","regression":"TestNativeNameBeforeSDNDeltaReachesMonitoring, TestUKNonLatinNamesAreProbes, TestUKChangesForUIDs, TestAdminReverseScreenUK, TestEqualComparesNonLatinNames"}},{"date":"2026-10-14","category":"fixed","summary":"POST /v1/screen no longer slows down or returns 504 while portfolio monitoring re-screens a large list change","details":"When a list change or a matching-engine update is screened against monitored portfolios, the monitoring job records its progress as it goes. Each progress write ran an integrity check that guards the job's audit record, and on a large job that check cost far more than the write it guarded. During a large re-screen on 2026-10-13 those checks crowded out live screening: POST /v1/screen ran several times slower, and some requests timed out with 504. The check now costs a small fraction of what it did. Its rules are unchanged: a production process still cannot alter a job's recorded facts (what was screened, against which list version) or delete the record, and a field added to the record later is protected automatically. No request or response field changes.","endpoint":"/v1/screen","verification":{"steps":["for i in $(seq 1 30); do curl -s -o /dev/null -w '%{http_code} %{time_total}\\n' -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}'; sleep 10; done | awk '{c[$1]++; if($2\u003em)m=$2} END{for(k in c) print k, c[k]; print \"max_s\", m}'","curl -s \"$BASE/v1/lists\" | jq -c '[.lists[] | select(.monitoring != \"ok\")]'"],"expect":"run the loop while a list change is being screened against portfolios (GET /v1/lists shows a recent update): it prints \"200 30\" and a max_s under 2; the second command prints []","audit":"monitoring job records keep their protection: what a job screened and against which list version cannot be altered or deleted by the production service, including any field added later","regression":"TestFactImmutabilityCost, TestAuditImmutability"}},{"date":"2026-10-14","category":"fixed","summary":"A re-screen after a matching-rules change no longer re-raises alerts your analysts already closed, unless the new match is stronger; alerts it does raise on a closed pair say so in prior_disposition","details":"After a change to Noble's matching rules, every monitored portfolio is re-screened (alerts from that carry change_type engine_change). Until now that re-screen treated only an open alert as already covering a pair, so a pair your analyst had closed as a false positive came back as a new alert after every rules change, at the same score and with nothing new to decide. Now, for the same portfolio entity, listed party and list: a pair closed as a true match is never re-alerted by a rules re-screen; a pair closed as a false positive is re-alerted only when the new score is higher than the score it was dismissed at, or the match came through a match pathway the dismissed alert's evidence did not include. When it is re-alerted, the new alert carries prior_disposition: the earlier alert's id, its status and score, and realerted_because (higher_score or new_pathway). When it is not, no alert is raised and the closed alert's activity timeline (GET /v1/alerts/{id}) gets a list_event entry recording that the re-screen reached the pair, at what score. A publisher's own change to a listing still alerts as before, whatever the earlier disposition: that is new information about the party. prior_disposition is an additive field on GET /v1/alerts and GET /v1/alerts/{id}; nothing else changes shape.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"$BASE/v1/alerts?source=monitoring\u0026limit=100\" -H \"X-API-Key: $KEY\" | jq -c '[.data[] | select(.prior_disposition) | {id, score, prior_disposition}]'","curl -s \"$BASE/v1/alerts/$CLOSED_ID\" -H \"X-API-Key: $KEY\" | jq -c '[.activity[] | select(.action == \"list_event\") | .note]'"],"expect":"after a rules re-screen, every printed alert has a score above its prior_disposition.score with realerted_because \"higher_score\", or realerted_because \"new_pathway\"; no alert follows a pair closed as a true match. For a monitoring alert you closed as a false positive before the re-screen ($CLOSED_ID), the second command prints a note beginning \"engine re-screen reached this closed_false_positive pair and did not re-alert\" unless the pair was re-alerted","audit":"every rules re-screen that reaches a closed pair leaves a record: either a new alert naming the closed one and why it was raised again, or an entry on the closed alert's timeline","regression":"TestEngineRescreenRespectsDispositions, TestGetAlertDecodesPriorDisposition"}},{"date":"2026-10-13","category":"fixed","summary":"Portfolio monitoring raises one alert per listed party, not one per matching alias","details":"When a sanctions list change matched a portfolio name, monitoring opened a separate alert for every one of the listed party's names that matched it: its primary name, each alias and each AI name variation. One party could fill a queue with three or four alerts, each needing its own review and four-eyes sign-off, and each sent its own webhook. Now a portfolio name matched against one listed party on one list is one open alert. Every further name of that party that matches is recorded on the alert's timeline as a list_event, with the name that matched and its score, and the alert is created from the strongest of the names screened together. A party listed on two lists is still two alerts, one per list, as before. Alerts already open are unchanged: a later match of the same party folds into one of them rather than opening another. No field changes on any endpoint.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"$BASE/v1/alerts?source=monitoring\u0026from=2026-10-14\u0026limit=100\" -H \"X-API-Key: $KEY\" | jq '[.data[] | select(.sdn_uid != null) | {e: .screened_name, uid: .sdn_uid}] | group_by(.e, .uid) | map(length) | max // 0'"],"expect":"0 or 1: no portfolio name holds two monitoring alerts for the same OFAC party created on or after this release (requires the portfolio monitoring add-on); a further match of the same party appears as a list_event in GET /v1/alerts/{id} activity","audit":"every name of the party that matched is kept on the alert's immutable timeline with its score and the list change that produced it, so the review record shows all the evidence on one alert","regression":"TestReverseJobOneAlertPerDesignation, TestMonitoringDedupKeysOnDesignation, TestMonitoringDedupMatchesPreUIDAlert, TestCrossListDesignationCreatesDistinctAlert"}},{"date":"2026-10-11","category":"added","summary":"GET /v1/export records list the alerts each screening raised, including the already-open alert a repeat screening folded into","details":"A screening's export record now carries an alerts array: every alert the screening created, and every alert it matched that was already open, whose id its response returned in place of a new one. Each entry gives the alert's id, its source, and its relation (created or deduplicated). An empty array means the screening raised no alert. Until now the record carried neither, so it could not show on its own whether a screening led to an alert, and a repeat screening of the same name within minutes was linked to the alert it returned nowhere in the record. Repeat screenings are linked from this release on; earlier ones were not recorded and cannot be reconstructed. Batch records return an empty array for now. Applies to GET /v1/export and GET /v1/export/bulk (JSON); the CSV format is unchanged. Additive: no existing field changes.","endpoint":"/v1/export","verification":{"steps":["A=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq -r '.trace_id + \" \" + (.alert_id|tostring)')","B=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq -r '.trace_id + \" \" + (.alert_id|tostring)')","echo \"$A | $B\"; for T in ${A%% *} ${B%% *}; do curl -s \"$BASE/v1/export?trace_id=$T\" -H \"X-API-Key: $KEY\" | jq -c .alerts; done"],"expect":"both screenings print the same alert_id; the first export prints [{\"id\":\u003cthat id\u003e,\"source\":\"api\",\"relation\":\"created\"}] and the second [{\"id\":\u003cthat id\u003e,\"source\":\"api\",\"relation\":\"deduplicated\"}] (Standard tier or higher, which export requires; run both screenings within five minutes)","audit":"the repeat screening's link to the alert is stored append-only when the screening runs, so the record shows the alert it returned without relying on the response the caller kept","regression":"TestExportListsScreeningAlerts, TestGetScreeningDecodesSingleRecord"}},{"date":"2026-10-11","category":"fixed","summary":"A sanctions list change can no longer miss portfolio monitoring; GET /v1/lists adds a monitoring field that says whether every imported change has been queued for your portfolio","details":"When a sanctions list changes, Noble imports the change and screens it against every monitored portfolio. Until now those were two separate steps: if the second one failed, or a deploy or restart landed between them, the change was imported but never screened against portfolios, and nothing retried it, so a customer matching a new designation could miss the monitoring alert and webhook. Now each list change is recorded for portfolio monitoring in the same step that imports it, and Noble queues it automatically, right after the import, when the service starts, and every minute until it succeeds. A failure or a deploy can delay a change's monitoring alerts; it can no longer lose them. This also covers names OFAC adds in a party's own script. GET /v1/lists gains one field per list, monitoring: \"ok\" when every change Noble has imported from that list has been queued for portfolio monitoring, \"delayed\" when a change has waited more than 15 minutes, which Noble alarms on and keeps retrying. Screening with POST /v1/screen is unaffected. This is an additive field: existing fields are unchanged.","endpoint":"/v1/lists","verification":{"steps":["curl -s \"$BASE/v1/lists\" | jq -c '[.lists[] | {list, status, monitoring}]'"],"expect":"every list carries a monitoring value of \"ok\" or \"delayed\"; on a healthy deployment every list prints monitoring \"ok\"","audit":"each list change keeps a durable record of when it was imported and which portfolio-monitoring job it became, so an examiner can see that every change was screened against portfolios and when","regression":"TestWatcherDeltaAlwaysQueued, TestListsReportDelayedMonitoring, TestMonitoringContractMatchesOutbox"}},{"date":"2026-10-11","category":"fixed","summary":"Alert, case-activity and portfolio listings reject a malformed filter with 400 instead of silently ignoring it, and a failed count or audit-trail lookup is an error instead of a quiet zero","details":"A filter that could not be read used to be dropped without a word: GET /v1/alerts?min_score=abc or ?from=2026/01/01 returned the whole queue with 200, so a filtered-looking view was not filtered. The same applied to the alert_id, from and to filters on GET /v1/alerts/activity and /v1/alerts/activity/stats, and to min_score and max_score on GET /v1/portfolio and /v1/portfolio/stats. Now every filter you send either applies or the request fails with 400 invalid_parameter, and param names the filter. Scores must be numbers from 0 to 100 (fractions allowed), with min_score no greater than max_score; dates must be YYYY-MM-DD, with from no later than to; alert_id must be a positive integer; status and source on GET /v1/alerts must be one of their documented values. A score bound of 0 now applies like any other: max_score=0 on the portfolio means a score of 0, where it used to mean no upper bound. Separately, a database error while building an alert's activity timeline, the GET /v1/alerts/stats counts, a list or export page, or an OFAC report's resolution details now returns 500 rather than a response with the missing part quietly empty or zero, so an incomplete audit trail is never presented as a complete one.","endpoint":"/v1/alerts","verification":{"steps":["curl -s -o /dev/null -w '%{http_code}\\n' \"$BASE/v1/alerts?min_score=abc\" -H \"X-API-Key: $KEY\"","curl -s \"$BASE/v1/alerts?from=2026/01/01\" -H \"X-API-Key: $KEY\" | jq -c '{code, param}'","curl -s \"$BASE/v1/alerts/activity?alert_id=abc\" -H \"X-API-Key: $KEY\" | jq -c '{code, param}'","curl -s \"$BASE/v1/portfolio?max_score=x\" -H \"X-API-Key: $KEY\" | jq -c '{code, param}'"],"expect":"the first prints 400; the others print {\"code\":\"invalid_parameter\",\"param\":\"from\"}, {\"code\":\"invalid_parameter\",\"param\":\"alert_id\"} and {\"code\":\"invalid_parameter\",\"param\":\"max_score\"}","audit":"a filtered alert, activity or portfolio view is now exactly the filter requested, or an error naming the filter that could not be applied","regression":"TestGETFiltersRejectMalformed, TestFilterParsers"}},{"date":"2026-10-11","category":"fixed","summary":"Monitoring alerts for names added by POST /v1/screen or POST /v1/portfolio now appear in GET /v1/alerts and are delivered by webhook","details":"A name enrolled in portfolio monitoring by screening it (with the portfolio add-on) or by POST /v1/portfolio was stored without its link to your account's workspace, and so was every monitoring alert it later raised. Those alerts were recorded but did not appear in GET /v1/alerts or GET /v1/alerts/{id}, were not delivered to your webhook, and were judged against the default alert threshold rather than yours. Names enrolled through POST /v1/batch were not affected. Both endpoints now attach every enrolled name to your workspace, and an enrollment that cannot be attached fails with a 500 instead of being stored unattached. Names enrolled before this fix, and the alerts they already raised, have been attached to your workspace, so those alerts now appear in GET /v1/alerts with their original created_at; they were not delivered by webhook when they were raised and are not re-sent now. No match, score or response field changes.","endpoint":"/v1/alerts","verification":{"steps":["curl -s -X POST \"$BASE/v1/portfolio\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"Ramzan Kadyrov\"}' | jq -c '{id, full_name}'","curl -s \"$BASE/v1/alerts?source=monitoring\u0026limit=100\" -H \"X-API-Key: $KEY\" | jq -c '[.data[] | {id, screened_name, source}]'"],"expect":"the first command returns 201 with an id; after the next sanctions-list change that matches a name in your portfolio, the second lists a monitoring alert for it whatever endpoint enrolled the name, and a registered webhook receives one alert.created event with the same id","audit":"each monitoring alert keeps its portfolio entity and trace_id, and GET /v1/alerts/{id} returns it to the account that enrolled the name","regression":"TestMonitoringAlertCarriesTenant"}},{"date":"2026-10-10","category":"added","summary":"scan_pin carries native_names_max_id, and replay reproduces matches made through names published in a party's own script","details":"The names OFAC publishes in a party's own script come from a separate OFAC publication that is imported on its own schedule, so they can be added after the OFAC list version a screening cites. Every screening that scans OFAC now records the highest of those names it could see, returned as scan_pin.native_names_max_id on GET /v1/export, and POST /v1/export/replay screens exactly those names again. Replay is no longer limited to the ones that existed when the cited OFAC list version was imported. A screening matched through ofac_native_script now replays with verdict match. Records stored before this field existed omit it and replay as before. No match, score or alert changes on POST /v1/screen.","endpoint":"/v1/export/replay","verification":{"steps":["T=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Путин Владимир\",\"lists\":[\"ofac\"]}' | jq -r .trace_id)","curl -s \"$BASE/v1/export?trace_id=$T\" -H \"X-API-Key: $KEY\" | jq .scan_pin.native_names_max_id","curl -s -X POST \"$BASE/v1/export/replay\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"trace_id\\\":\\\"$T\\\"}\" | jq -c '{verdict, differences}'"],"expect":"the second command prints a number of at least 1; the third prints verdict \"match\" and no differences (Standard tier or higher, which replay requires)","audit":"the ceiling is stored with the screening's other list-version identities, so a replay cites exactly the native-script names the original scan could see","regression":"TestReplayReproducesNativeNameAddedAfterPublish, TestNativeNamesCutOnTheirOwnID"}},{"date":"2026-10-10","category":"added","summary":"Portfolio entities carry name_fold: the form of the name that monitoring compares against the lists","details":"Every entity returned by GET /v1/portfolio, POST /v1/portfolio and PUT /v1/portfolio now has name_fold: the name in lower case with accents removed (Ališer Usmanov becomes aliser usmanov), ligatures and fullwidth letters expanded, apostrophes dropped, punctuation as spaces, and lookalike letters from another script inside a Latin word mapped to Latin. It is the same fold every list name goes through, so you can see exactly what your customer's name is compared as. It is written when an entity is enrolled or renamed; entities enrolled before today show it once the portfolio has been re-folded, and omit it until then. For a Confidential Portfolio it is computed from the decrypted name when you read the entity and is never stored. This release stores and returns the fold only: monitoring does not use it yet, and no match, score or alert changes. Monitoring will compare on it in a later release, announced here with its own engine_version.","endpoint":"/v1/portfolio","verification":{"steps":["curl -s -X POST \"$BASE/v1/portfolio\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"full_name\":\"Ališer Usmanov\"}' | jq -c '{id, full_name, name_fold}'"],"expect":"prints full_name \"Ališer Usmanov\" and name_fold \"aliser usmanov\"; the same id listed by GET /v1/portfolio shows the same name_fold","audit":"the fold is stored beside the name it was computed from, with the version of the folding rules that produced it, so a later rules change re-folds exactly the rows it affects","regression":"TestPortfolioNameFold, TestPortfolioListReturnsNameFold, TestFold"}},{"date":"2026-10-10","category":"changed","summary":"Names OFAC publishes in a party's own script — 8,262 Cyrillic, Arabic, Chinese, Korean and other names across 4,844 parties — are now screened; engine_version 1.2.0; the script_not_screened warning now says exactly what was not compared","details":"OFAC publishes its list in two formats. The one Noble imported is Latin-only. The other, the SDN Advanced format, also carries the names OFAC documents in a party's own script: Путин Владимир Владимирович, سليم عبد الله سليم الزق and thousands more. Those names are now imported from it and screened as their own pathway, ofac_native_script, with the matched form in variation and a reason naming the script; each multi-part name is matched written given-name first or surname-first, so either convention scores 100. Which parties are listed still comes from the main OFAC list, so the two formats never disagree about that; a native name published before its party reaches the main list starts matching once it does. Native names also take part in portfolio monitoring: a party that gains one is screened against monitored portfolios as if its listing had changed. This is a change to what the engine finds, so engine_version becomes 1.2.0 and monitored portfolios are re-screened. SECOND, the script_not_screened warning was worded for a world in which OFAC published Latin only. Its code and when it appears are unchanged, but the message now says what is true: the non-Latin part of a name was compared against the names the lists publish in that script (OFAC, the UN and the UK publish them for some parties) and was not romanized, so a party listed only in Latin script was not compared against it. summary.text and the MCP narration say the same. WHAT YOU WILL SEE: matches with source ofac_native_script for names OFAC publishes in their own script, an exact native form of such a name scoring 100 and alerting, engine_version \"1.2.0\", and reworded warning messages. No field changed shape and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["for n in 'سليم عبد الله سليم الزق' 'Владимир Владимирович Путин' 'Путин Владимир'; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n,lists:[\"ofac\"],limit:5}')\" | jq -c --arg n \"$n\" '{n:$n, engine_version, uid:.summary.top_match.uid, score:.summary.top_match.score, alert:(.alert_id!=null), via:[.matches[0].match_sources[]|select(.source==\"ofac_native_script\")|.reason], warn:[.warnings[]?.message][0][0:80]}'; done"],"expect":"every line prints engine_version \"1.2.0\" and an ofac_native_script reason naming the script. The Arabic line prints uid \"57717\", score 100, alert true; the two Cyrillic lines print uid \"35096\", alert true, the full name at score 100. Each warn begins \"Part of this name is written in \u003cscript\u003e script. Noble compared that part only against names\"","audit":"every native name is stored with the uid it belongs to, its script and the date of the OFAC publication it came from, and every match through one cites the form and script in the stored screening; native-name refreshes that reach monitored portfolios run as ordinary reverse-screen jobs","regression":"TestParseAdvancedNativeNames, TestOFACNativeNamesMatch, TestOFACNativeNamesReachMonitoring, TestNameFormsCorpus, TestScreenWarnings, TestSummarizeScreenNarratesUnscreenedScript"}},{"date":"2026-10-10","category":"fixed","summary":"A name written in the script the UN publishes it in — Arabic, Cyrillic, Persian and others — now matches its UN designation; it was dropped at import","details":"The UN Security Council list publishes 377 parties' names in their original script as well as in Latin (the NAME_ORIGINAL_SCRIPT field), for example صدام حسين التكريتي for Saddam Hussein Al-Tikriti. Noble's UN importer did not read that field, so screening the published original-script name against lists:[\"un\"] returned no match at all. It is now imported and screened. A match through it is reported as its own source, un_original_script, with the matched native-script form in variation and a reason saying it is the name the UN publishes in the party's own script — never as an AI guess. These names also take part in portfolio monitoring: the first UN list update after this change reports the affected designations as modified, because each gained a published name, and monitored portfolios are screened against the native forms then. WHAT YOU WILL SEE: matches for UN-listed parties screened by their published original-script name, with source un_original_script. No field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"صدام حسين التكريتي\",\"lists\":[\"un\"],\"limit\":5}' | jq -c '{uid:.summary.top_match.uid, score:.summary.top_match.score, alert:(.alert_id!=null), via:[.matches[0].match_sources[]|select(.source==\"un_original_script\")|{variation,reason}]}'"],"expect":"prints uid \"UN-6908048\", score 100, alert true, and one un_original_script source whose variation is the Arabic name and whose reason says it is the name published in its original script. The Latin form Saddam Hussein Al-Tikriti still returns UN-6908048 at 100","audit":"the native-script name is stored with the designation it belongs to and cited in the match's source detail, so the stored screening shows exactly which published form matched","regression":"TestUNOriginalScriptNamesMatch, TestUNFreshLoadDeltaRoundTrip, TestRegistryCTEsMatchAuditedSQL, TestLiveQueryMatchesGolden"}},{"date":"2026-10-09","category":"changed","summary":"AI name variations now cover the OFAC Consolidated list and include native-script forms and the renderings used by non-English press; engine_version 1.1.0","details":"Noble screens every name against pre-computed variations of each OFAC listing — transliterations, spelling variants and cultural equivalents, each stored with a reason. Until now those variations existed for the SDN list only and were Latin-script only, so a name sent in its native script (Arabic, Persian, Cyrillic, Hangul, Han) or in the spelling a German, French, Spanish, Polish, Turkish or Italian newspaper uses could score low against a party OFAC lists in English, and a party on the Consolidated list had no variations at all. Two changes. FIRST, the variation set is broadened: for each listing it now asks for the party's native-script forms, every standard romanization system, the non-English press renderings, common misspellings and cultural equivalents, capped per name, each with a reason naming the system or language and a version of the prompt that produced it. Existing listings are regenerated under the new prompt, keeping what they had and adding only new forms; the backfill runs in-cluster and is announced here when complete. SECOND, the Consolidated list is covered by the same pathway, which is a change to what the engine can find, so engine_version becomes 1.1.0 and every monitored portfolio is re-screened against the new forms as they land (alerts from that carry change_type engine_change). WHAT YOU WILL SEE: higher scores and new matches for native-script and foreign-press forms of OFAC-listed names, with match sources of type ofac_ai_variation whose reason names the form; new matches against Consolidated-list parties through the same source; engine_version \"1.1.0\" on responses. Scores of names that already matched well do not fall. No field changed shape and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["for n in 'قاسم سليماني' '김정은' 'Sergei Lawrow'; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n,lists:[\"ofac\"],limit:5}')\" | jq -c --arg n \"$n\" '{n:$n, engine_version, uid:.summary.top_match.uid, score:.summary.top_match.score, alert:(.alert_id!=null), ai:[.matches[0].match_sources[]?|select(.source==\"ofac_ai_variation\")|.reason]}'; done"],"expect":"every line prints engine_version \"1.1.0\". Once the backfill announced here has completed: قاسم سليماني prints uid 10481, alert true and an ai reason naming the Persian or Arabic script form; 김정은 prints uid 20157, alert true and a Hangul reason; Sergei Lawrow prints uid 35233, alert true and a reason naming the German rendering. Before the backfill completes the three may still miss; the engine_version line is the part this entry promises on deploy","audit":"every variation row carries the reason the model gave, the prompt version that produced it and the OFAC list it belongs to, and every match through one names the variation that matched; the backfill's reverse-screen jobs and any engine_change alerts are recorded with the engine_version","regression":"TestVariationsBackfillBroadened, TestVariationsBackfillDryRunCallsNothing, TestBatchPromptAsksForEveryFormClass, TestBatchPromptExcludesExistingForms, TestReverseScreenAIVariationParity"}},{"date":"2026-10-09","category":"added","summary":"After a change to Noble's matching rules, every monitored portfolio is re-screened against every listed name; alerts raised that way carry change_type engine_change","details":"Portfolio monitoring reacts to list changes: when a publisher adds or amends a designation, the changed name is screened against every monitored portfolio. A change to Noble's own matching rules changes nothing on any list, so until now a rule that finds more left every monitored entity with the clear it earned under the old rules. Now, after a rules change (a new engine_version), Noble re-screens every currently listed name on every list against every monitored portfolio, through the same monitoring path a list change uses, with the same alert threshold and the same evidence. An alert raised this way carries change_type \"engine_change\" and names the engine_version in its screening record, so it is distinguishable from an alert caused by a publisher's change. A name that already has an open alert is not alerted twice. WHAT YOU WILL SEE: after a rules change, possibly new alerts on GET /v1/alerts with change_type \"engine_change\", each with the usual evidence; nothing otherwise changes. The rules change itself is always announced here first. No field changed meaning and no version is minted.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"$BASE/v1/alerts?limit=5\" -H \"X-API-Key: $KEY\" | jq -c '.data[] | {id, change_type, screened_name, score}'"],"expect":"lists recent alerts; after a rules change that found a new match in your portfolio, one or more carry change_type \"engine_change\". Before any rules change, none do. Operator evidence for each rules change: the re-screen's job count, total names and engine_change alert count, recorded on the change's issue","audit":"each engine_change alert links to a monitoring screening record carrying the engine_version that warranted it and the list versions it was screened against, so an examiner can see that a rules change was followed by a full re-screen and what it found","regression":"TestAdminRescreen, TestChunkChangesRespectsNameCount, TestChunkChangesOversizeDesignationIsOwnChunk"}},{"date":"2026-10-09","category":"added","summary":"Every screening now names the matching rules that produced it — engine_version and match_engine — and replay tells a rules change apart from data drift","details":"Noble's matching rules change: a transliteration class gets closed, a scoring penalty gets calibrated. Until now nothing recorded which rules produced a decision, so a replay of a screening made under earlier rules re-ran today's rules and reported the difference as if the lists had drifted. Two fields close that. engine_version names the rules in force; it changes only with a rules change, and every such change ships with a changelog entry. match_engine names the engine that scored the candidates: \"database\" today. Both appear on POST /v1/screen and on each item of GET /v1/batch/status, are stored with the record, and are reported by the MCP screen_name tool. POST /v1/export/replay compares the record's engine_version to the current one: on a mismatch it adds the engine_version_changed caveat and stops comparing scores, counts and the summary verdict, while still comparing which records matched and which list versions were read. Records stored before these fields existed carry no version, are reported without the fields, and replay in full as before. WHAT YOU WILL SEE: two new strings on every new screening response and batch item, and a possible new caveat on replay. Additive: no field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq -c '{engine_version, match_engine}'","curl -s \"$BASE/v1/changelog\" | jq -r '.entries[0].summary'"],"expect":"step 1 prints a non-empty engine_version and match_engine \"database\"; step 2 prints this entry's summary. A replay of a record stored before this change prints no engine_version_changed caveat; a replay of a record stored under an earlier engine_version prints it and no score differences","audit":"engine_version and match_engine are stored on the screening record (engine_version, match_engine columns) and appear in the stored response payload, so an examiner can see which rules produced any decision and a portfolio re-screen can be keyed to the change that warrants it","regression":"TestScreeningReportsAndStoresEngineVersion, TestReplayOfOlderEngineVersionCarriesCaveatNotScoreDrift, TestScreenDecodesEngineFields, TestBatchDecodesEngineFields, TestObserveRunsCandidateOffPathAndReports"}},{"date":"2026-10-09","category":"fixed","summary":"list_versions is now read inside the screening's own database transaction, so a response can never cite a list version the scan did not see, and replay is keyed on those exact versions rather than on the record's clock","details":"A name screening reads its candidates in one read-only database transaction, so every part of the scan sees the same list snapshot. The list_versions it reported were looked up afterwards, in separate statements, and the replay endpoint re-ran a recorded screening against the lists as they stood at the record's own timestamp. Both happened after the scan. A list import committing in the gap could make a response cite a publish id the scan had not read, and could make a replay read a different list state than the original had, reporting a difference where no decision had changed. Both gaps are closed. The list versions are now read inside the scan transaction, under the snapshot the candidates came from. Every screening record stores what its scan read, by id: the publish id of each list and the highest AI-variation id visible. POST /v1/export/replay rebuilds its list state from those ids, list by list, and from the recorded variation ceiling, so it reproduces exactly what the original scan read however imports interleaved with it. A record stored before this change has no such pin; its replay falls back to the record's timestamp as before and says so with a new caveat, pinned_to_record_time. WHAT YOU WILL SEE: no change to any response shape. list_versions on POST /v1/screen carries a stronger guarantee; replay responses for older records may carry the pinned_to_record_time caveat, and replays of records stored from now on do not. No version is minted.","endpoint":"/v1/export/replay","verification":{"steps":["R=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}'); T=$(echo \"$R\" | jq -r .trace_id); echo \"$R\" | jq -c '[.list_versions[]|{list,publish_id}]'","curl -s -X POST \"$BASE/v1/export/replay\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"trace_id\\\":\\\"$T\\\"}\" | jq -c '{verdict, caveats, lv:[.replayed.list_versions[]|{list,publish_id}]}'"],"expect":"step 2 prints verdict \"match\", a caveats array that does not contain \"pinned_to_record_time\" (it may contain the OFAC child-detail caveat), and a list_versions array identical to step 1's. A replay of a record stored before this change prints the pinned_to_record_time caveat and, absent a mid-scan import, still matches","audit":"every new screening record carries the instant of its scan transaction and the publish ids it read (scan_tx_at, scan_vector), so an examiner can see not only which list versions a decision cites but that the engine read exactly those versions; records predating the change are unchanged and distinguishable by the caveat","regression":"TestScanPinStoredWithScreening, TestReplayOfPinnedRecordMatchesWithoutCaveat, TestReplayOfLegacyRecordCarriesCaveat, TestCutAtPublishesPinsEachListSeparately, TestImportStampsAliasPublishID"}},{"date":"2026-10-09","category":"added","summary":"The Server-Timing header on a name screening now breaks the server's time into stages — scan, score, enrich, persist — so a slow screening shows where the time went","details":"Every /v1/ response has carried Server-Timing: total;dur=\u003cms\u003e, the server's own measurement of the request. A single number says a screening was slow but not why. POST /v1/screen now adds one metric per pipeline stage that ran, in pipeline order: scan (the candidate search across the sanctions lists, your watchlist, identifier probes and deep-screen name forms), score (scoring and ranking the candidates, attribute comparison), enrich (match enrichment, ownership analysis, the list-version lookup) and persist (the audit-trail write). Stages are a breakdown of total, not additions to it, and need not sum to it: request parsing, alerting and the response write are in total only. A stage is present only when it ran. These four names are stable under the current version: a stage may be added, never renamed or removed. Every other route carries total alone. WHAT YOU WILL SEE: Server-Timing on POST /v1/screen reads e.g. total;dur=412.5, scan;dur=247.1, score;dur=7.2, enrich;dur=61.0, persist;dur=5.3. Additive: the total metric is unchanged, no field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -sD - -o /dev/null -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | grep -i '^server-timing'","curl -sD - -o /dev/null \"$BASE/v1/status\" -H \"X-API-Key: $KEY\" | grep -i '^server-timing'"],"expect":"step 1 prints one Server-Timing header of the form total;dur=\u003cms\u003e, scan;dur=\u003cms\u003e, score;dur=\u003cms\u003e, enrich;dur=\u003cms\u003e, persist;dur=\u003cms\u003e, every metric present and total at least as large as scan; step 2 prints total;dur=\u003cms\u003e and nothing else","audit":"the same stage timings are written to the request's structured log line under its trace_id and request_id, so an operator can reconcile a customer's measurement against the server's record","regression":"TestScreenServerTimingReportsStages, TestServerTimingReportsRecordedStages, TestServerTimingCarriesOnlyTotal, TestPublicStageCoversEveryTimedPhase, TestStageTimingsRenderInPipelineOrder"}},{"date":"2026-10-09","category":"fixed","summary":"EU names published only as a whole name — most entities, and every native-script alias — are now screened; before this they could not be reached by any query","details":"The EU consolidated list publishes a name either as a first and last name or as a single whole name. Entities (companies, groups, vessels) and native-script aliases are almost always the second kind. Noble's searchable EU name was built from the first and last name only, so a party published with a whole name alone had an empty search string and no query could reach it: roughly half of EU designations and three quarters of EU aliases, Sberbank and the Wagner Group among them. Screening 'Sberbank' against lists:[\"eu\"] returned unrelated people and no alert. On a default screen the same party was usually caught through OFAC or the UK list, which is why this went unnoticed; a customer screening lists:[\"eu\"] alone, or a party the EU lists and no one else does, had a miss. Every EU name now searches on its published whole name when no first and last name exist. Names that had a first and last name are unchanged, so no existing score moved. A whole-name match now displays the published name where it previously showed blank. WHAT YOU WILL SEE: new EU matches and alerts for names that previously returned nothing against the EU list, including native-script forms such as Сбербанк. No field changed meaning and no version is minted. If you monitor a portfolio, entities listed by the EU alone may raise alerts on the next EU list update.","endpoint":"/v1/screen","verification":{"steps":["for n in 'Sberbank' 'Wagner Group' 'Сбербанк' 'Oleg Deripaska'; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n,lists:[\"eu\"],limit:5}')\" | jq -c --arg n \"$n\" '{n:$n, top:(.summary.top_match|{uid,name,score}), alert:(.alert_id!=null)}'; done"],"expect":"Sberbank prints top {\"uid\":\"EU-142064\",\"name\":\"Sberbank\",\"score\":100} and alert true; Wagner Group prints a top uid of EU-132563 or EU-152064 (two EU parties publish that alias) at score 100 and alert true; Сбербанк prints top uid EU-142064 and alert true; Oleg Deripaska prints top uid EU-138212 at score 100 and alert true, unchanged from before (the original control, Alisher Usmanov, was delisted by the EU on 2026-10-12)","audit":"each screening is persisted with its trace_id and the EU publish_id it was scored against, so an export shows the match the caller saw; a screening of the same name recorded before this change carries the earlier result and replays with a difference, which is the recorded evidence of the gap","regression":"TestEUWholeNameRowsSearchable, TestEUWholeNameScreening, TestSearchColumnsAreGenerated, TestRegistryCTEsMatchAuditedSQL, TestLiveQueryMatchesGolden"}},{"date":"2026-10-09","category":"fixed","summary":"A UK or EU list update no longer reports a designation as modified, or re-screens your portfolio against it, when nothing about it changed","details":"When a list update arrives, Noble compares every designation it already holds against the new publication to decide what was added, removed or modified, and a modification re-screens every monitored portfolio against that designation. For the UK and EU lists that comparison could misread an unchanged designation as modified: two of its aliases that share a surname and first name — a name with and without a middle name, or the same name in two languages — were compared in whichever order the database happened to return them. The designation had not changed; its alias list had been read back in a different order. The effect was spurious modified entries in the list change history and unnecessary re-screening jobs after some UK and EU updates. The comparison now orders aliases on every field, so the same designation always compares equal to itself. A real change to any alias field is still a modification. WHAT YOU WILL SEE: fewer modified entries after a UK or EU list update, and no portfolio re-screen for a designation that did not change. Nothing in any response changed shape and no version is minted.","endpoint":"/v1/lists","verification":{"steps":["curl -s \"$BASE/v1/lists\" -H \"X-API-Key: $KEY\" | jq -c '.lists[] | select(.list==\"uk\" or .list==\"eu\") | {list, publish_id, record_count}'"],"expect":"both lists report their current publish_id and record_count as before; the fix changes what an update records, not what this endpoint returns. Operator evidence: after the next UK and EU publications, every row in uk.designation_changelog and eu.designation_changelog with change_type = modified for that publish_id has old_data and new_data that differ once aliases are compared as sets — zero rows where they are identical","audit":"the list change history is append-only; entries recorded before this fix are unchanged, and any spurious modified entry already recorded remains visible with its publish_id","regression":"TestEqualIgnoresAliasOrderOnTiedKeys (uk and eu), TestUKFreshLoadDeltaRoundTrip"}},{"date":"2026-10-08","category":"added","summary":"A name written in a non-Latin script — Cyrillic, Arabic, Han, Hangul and others — now carries a script_not_screened warning, so a low or empty result for it can no longer read as clear","details":"Most sanctions lists, OFAC among them, publish names only in Latin script, and Noble does not yet romanize a name before comparing it. A name submitted in its original script, such as Владимир Путин, أسامة بن لادن or 김정은, was compared only against the lists that publish names in that script. The response could come back with low scores or no matches and nothing to say why: a confident all-clear for a name that had never been compared against OFAC. That silence is closed. Every screening response now has an optional warnings array. When part of the name is written in a non-Latin script, it carries one entry per script: {\"code\":\"script_not_screened\",\"script\":\"Cyrillic\",\"message\":\"...\"}. summary.text ends with the same caveat, and the warning is stored with the screening, so an export shows the caveat the caller saw. Batch items carry the same warnings, and the MCP screen_name tool states the caveat in its narrated result. A name written wholly in Latin script, accented letters included, carries no warning and is unchanged. Until romanization ships, screen the romanized form of the name as well. WHAT YOU WILL SEE: an optional warnings array on POST /v1/screen and on each item of GET /v1/batch/status, present only for names written partly or wholly in a non-Latin script. Additive: no field changed meaning and no version is minted. Treat an unrecognized warning code as a coverage caveat, never as clear.","endpoint":"/v1/screen","verification":{"steps":["for n in 'Владимир Путин' 'أسامة بن لادن' '김정은' 'Sergey Lavrov'; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"$(jq -cn --arg n \"$n\" '{name:$n}')\" | jq -c --arg n \"$n\" '{n:$n, warnings:[.warnings[]? | {code, script}], text:.summary.text}'; done"],"expect":"the Cyrillic, Arabic and Hangul names each print warnings [{\"code\":\"script_not_screened\",\"script\":\"Cyrillic\"}], [... \"Arabic\"] and [... \"Hangul\"] respectively, and each text ends with \"Not screened in \u003cscript\u003e script against lists that publish only Latin names, including OFAC — see warnings.\". Sergey Lavrov prints warnings [] and its text carries no caveat","audit":"the warning is persisted in the stored screening response and in summary.text, so GET /v1/export for the trace_id shows the same caveat the caller received","regression":"TestNonLatin, TestScreenWarnings, TestWithWarningProse, TestScreenResponseCarriesScriptWarning, TestScreenDecodesWarnings, TestSummarizeScreenNarratesUnscreenedScript"}},{"date":"2026-10-06","category":"fixed","summary":"A screening now answers or fails within a published deadline — 10 seconds, 25 with deep_screen or deep_summary — and a screening that fails is still recorded in the audit trail, retrievable by the request_id in its error","details":"A name screening could stall: the connection stayed open with no status, no error body and no request_id, and nothing was recorded. The caller could not tell whether the name had been screened, and an examiner could not see that the control had been tried and failed. Both gaps are closed. FIRST, POST /v1/screen has a deadline: 10 seconds, or 25 seconds when deep_screen or deep_summary is requested. A screening that cannot finish in time ends with HTTP 504 and the standard error body: type api_error, code screening_timeout, with a request_id and a trace_id. It never holds the connection open. No partial result is ever returned; a result assembled after the deadline is treated as a timeout. SECOND, a screening that fails after its input was accepted — a timeout, any server-side error, or a caller who hangs up — is written to the audit trail as a failed screening. GET /v1/export?request_id= (or ?trace_id=) returns it. Every export record now carries outcome: completed for a screening result, failed for an attempt that produced none, and a failed record's response is the exact error the caller received. CSV exports gain two trailing columns, outcome and failure_code, so a failed screening can no longer pass for a zero-match row. A failed screening has no result: screen the name again. POST /v1/export/replay refuses one with replay_unsupported_outcome. THIRD, the language-model calls behind deep_screen and deep_summary are now bounded. One that runs too long degrades exactly as a failed call already did: the screening answers with deep_screen_fallback_reason or summary.fallback_reason set to llm_failure. WHAT YOU WILL SEE: an outcome field on every export record, two new trailing CSV columns, an optional trace_id on 504 screening_timeout errors, and a 504 in place of an open connection. Set your client timeout a few seconds above the deadline that applies. Additive — no existing field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -o /tmp/screen.json -w '%{http_code} %{time_total}\\n' --max-time 30 -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladmir Pootin\"}' \u0026\u0026 jq -r '.trace_id // .request_id' /tmp/screen.json","curl -s \"$BASE/v1/export?trace_id=$(jq -r .trace_id /tmp/screen.json)\" -H \"X-API-Key: $KEY\" | jq -c '{outcome, screened_name}'","curl -s https://noblesight.io/llms.txt | grep -i 'screening deadline'"],"expect":"step 1 prints a status and a time no greater than 10 seconds. The status is 200, or 504 with a body whose code is screening_timeout and which carries request_id and trace_id; never a curl timeout with no status. Step 2 returns the record for that call: {\"outcome\":\"completed\",\"screened_name\":\"Vladmir Pootin\"} after a 200, or outcome \"failed\" after a 504. Step 3 prints the published deadline line","audit":"every accepted screening leaves an audit record whether or not it produced a result, under the request_id and trace_id returned to the caller, so an examiner can see that a screening was attempted and that it failed","regression":"TestScreenStallTimesOutAndIsRecorded, TestScreenClientDisconnectIsRecorded, TestTimedOutScreeningIsRetrievable, TestDeepScreenHungLLMDegrades, TestScreenDeadlineFor, TestGetScreeningDecodesFailedOutcome"}},{"date":"2026-10-05","category":"changed","summary":"The API now runs on more than one server instance, so routine infrastructure maintenance no longer interrupts screening; the signup rate limit is now documented as approximate","details":"Until now a single server instance answered every request, so planned infrastructure work on the machine it ran on briefly interrupted the API. Requests are now spread across more than one instance and maintenance takes them out of service one at a time, so screening stays available through it. One documented consequence: the signup endpoint's limit of 10 requests per minute per client IP is counted by each instance separately, so while several instances serve traffic the effective ceiling can be higher than 10. That limit exists to blunt signup abuse, and clients should not depend on its exact value. Daily and monthly API-key quotas are counted centrally and remain exact. WHAT YOU WILL SEE: nothing in any response changes. No field changed meaning and no version is minted.","endpoint":"/v1/keys","verification":{"steps":["for i in $(seq 40); do curl -s -o /dev/null -w '%{http_code}\\n' -X POST \"$BASE/v1/keys\" -H 'Content-Type: application/json' -d '{}'; done | sort | uniq -c","curl -s \"$BASE/v1/usage\" -H \"X-API-Key: $KEY\" | jq -c '{used, limit, remaining, window}'"],"expect":"step 1 shows at least one 429 among 40 rapid unauthenticated signup attempts from one IP: the limiter still engages, at a ceiling that may exceed 10 per minute. Step 2's quota counters are unchanged in meaning and keep counting every call exactly, whichever instance served it","audit":"every screening is still persisted with its trace_id and request_id, whichever instance served it, so the audit trail is unaffected","regression":"TestIPLimiter, TestIPLimiterIndependentIPs, TestIPLimiterWindowExpiry"}},{"date":"2026-09-23","category":"fixed","summary":"Name screening is several times faster on the server, and match_sources now comes back in a fixed order, strongest pathway first","details":"Most of the time a name screening spent inside our database went on reading every sanctions list in full rather than looking the name up. The lookups now do what they were built to do. Measured on the same data before and after, the server's own time for a typical nine-list name screening fell by a factor of three to nine; the gain is largest on multi-word names and smallest on single short tokens. Nothing about what is found changed: on a battery of name shapes the same records came back with the same scores and the same per-pathway detail, byte for byte. SECOND, match_sources is now ordered — strongest pathway first — and the choice between two equally close aliases of the same record is made on the aliases themselves rather than on the order our database happened to read them. Before, the array's order, and on an exact tie which alias was reported, could differ between two identical requests; now the same request against the same list state always returns the same array, and where a strong and a weak alias tie, the strong one is reported. WHAT YOU WILL SEE: a lower Server-Timing total on /v1/screen, and a stable match_sources order. No field was added or removed, no field changed meaning, and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["for i in 1 2 3 4 5; do curl -sD - -o /dev/null -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | grep -i '^server-timing'; sleep 5; done","for i in 1 2; do curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Mohammed\",\"limit\":100}' | jq -c '[.matches[] | {uid, score, s: [.match_sources[] | [.source, .trigram_score, .matched_text, .alias_quality]]}]' | md5sum; done"],"expect":"step 1 prints five Server-Timing headers whose median total;dur is under 600; step 2 prints the same checksum twice — identical matches, scores, and match_sources in identical order — and within each match the match_sources entries are in non-increasing trigram_score order","audit":"match_sources is persisted with the screening in the order the live response carried, so the stored record and the response a customer logged agree pathway for pathway","regression":"TestAddSourceIgnoresArrivalOrder, TestScreeningPredicatesAreIndexed, TestListTrigramIndexesHaveNoPendingList"}},{"date":"2026-09-16","category":"fixed","summary":"A listed name that contains a parenthetical — VEST SPECTRUM (S) PTE. LTD., NOOR ENERGY (MALAYSIA) LTD. — now scores 100 and raises an alert when screened exactly as the list publishes it","details":"Some sanctioned entities are published with a parenthetical in the name itself: a place of incorporation, an abbreviation, a legal form. Screening one of those names exactly as OFAC publishes it did not score 100 against its own record. The engine dropped the parenthetical from the name you submitted — and, by a defect, everything after it too — before comparing, while the listed name was compared in full, so the two could never agree. Forty-two SDN primary names and about sixty aliases were affected; on a sample of fourteen, six exact matches of a listed name raised no alert at all, and summary.top_match named an unrelated record. That is a false negative on the simplest possible input, and it is closed. The name you submit is now screened exactly as written. A submitted name that carries a parenthetical is additionally screened with the parenthetical removed, and each record keeps whichever comparison scored higher — so an annotation of your own in parentheses still reaches the record it reached before. WHAT YOU WILL SEE: an exact match of a published name that contains a parenthetical now returns the publishing pathway at 100, alerts, and names itself as summary.top_match. Names without a parenthetical are unaffected. Additive — no field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"VEST SPECTRUM (S) PTE. LTD.\",\"limit\":100}' | jq -c '{alert:(.alert_id!=null), top:.summary.top_match.uid, sdn:[.matches[]|select(.uid==\"11309\")|.match_sources[]|select(.source==\"ofac_sdn\")|.trigram_score]}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"WEE TIONG (S) PTE LTD\",\"limit\":100}' | jq -c '{alert:(.alert_id!=null), top:.summary.top_match.uid, sdn:[.matches[]|select(.uid==\"25687\")|.match_sources[]|select(.source==\"ofac_sdn\")|.trigram_score]}'"],"expect":"step 1 prints {\"alert\":true,\"top\":\"11309\",\"sdn\":[100]} and step 2 prints {\"alert\":true,\"top\":\"25687\",\"sdn\":[100]}: each exact match of a published name alerts, names its own record as the top match, and reports the ofac_sdn pathway at full similarity","audit":"the stored screening carries the name exactly as you submitted it, so an examiner replaying the record sees the same input and the same 100","regression":"TestExactNameParentheticalScores100, TestExactNameParentheticalTopMatchIsItself, TestExactNameParentheticalHTTP, TestParentheticalAnnotationStillReachesTheRecord, TestStripParentheticals, TestQueryForms, TestMergeFormsTakesBestPerRecordAndPerSource"}},{"date":"2026-09-15","category":"added","summary":"Every API response now carries a Server-Timing header reporting how long the server itself spent on the request, so you can measure our latency from your own client instead of taking our word for it","details":"Until now the only way to know how long a screening took was to time it yourself from the outside, which measures your network as much as our work. Every response under /v1/ now carries the standard Server-Timing header — total;dur=\u003cmilliseconds\u003e — holding the server's own measurement of that request. It is the figure our own monitoring records, not a separate estimate, and it is present on error responses too, which are the ones most worth timing when something is wrong. WHAT YOU WILL SEE: one additional response header, readable with curl -D- or from any HTTP client's header map, on every /v1/ endpoint including failures. Nothing in any response body changed, no field changed meaning, and no version is minted. A retried request served from an idempotency cache reports the time that retry actually took rather than replaying the original measurement. SECOND, an unexpected server fault now returns a structured error response carrying a request_id, in the same shape as every other error, rather than closing the connection.","endpoint":"/v1/screen","verification":{"steps":["curl -sD - -o /dev/null -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Ali Hassan\"}' | grep -i '^server-timing'","curl -sD - -o /dev/null \"$BASE/v1/lists\" | grep -i '^server-timing'"],"expect":"both steps print a Server-Timing header of the exact form total;dur=\u003cnumber\u003e, where the number is the milliseconds the server spent on that request and is greater than zero; the header carries the single total metric and nothing else","audit":"the figure is the server's own measurement of the same request the X-Request-Id header identifies, so a slow call can be quoted back to us by request_id with a latency we both agree on","regression":"TestServerTimingScopedToAPIPaths, TestServerTimingCarriesOnlyTotal, TestServerTimingPresentOnErrorResponses, TestRequestLogPanicBeforeWriteBecomesStructured500"}},{"date":"2026-09-15","category":"fixed","summary":"Portfolio monitoring now reaches the same alerting decision a live screening reaches: a monitored name one or two characters from a newly designated party raises an alert, and monitoring alerts carry the same explanation a screening carries","details":"Screening a name yourself and having us watch a name for you were not the same control, and the difference was in the direction that matters. A name you screen through the API that sits one or two characters from a published sanctions name raises an alert despite its score, because that is the shape a deliberate evasion takes. A name sitting in your monitored portfolio did not: when a list update designated a party it was that close to, the comparison happened, the score came out just under your threshold, and nothing was raised. That is the case portfolio monitoring exists for — someone screened clean when you onboarded them and designated later — so it was the wrong place to have the gap. It is closed: the delta path now runs the same scoring pipeline the live and batch paths run, and the alert carries alert_reason naming the rule, exactly as a live screening does. SECOND, monitoring alerts can now explain themselves. The stored evidence records how much of the score came from phonetic name agreement, and that figure was previously written as zero on every monitoring alert — including, by construction, the alerts where phonetic agreement was the thing that carried the match over the line. An analyst reading one of those was told the phonetic contribution was nothing when it was most of the score. WHAT YOU WILL SEE: a small number of additional monitoring alerts, each below your threshold and each carrying the signal naming the two tokens and the characters between them, and a real figure where there was a zero. Scores are unchanged, so any threshold you have tuned behaves as before. Additive — no existing field changed meaning and no version is minted.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"$BASE/v1/alerts?source=monitoring\u0026limit=20\" -H \"X-API-Key: $KEY\" | jq -c '[.data[] | {id, score, alert_reason, boost: .match_details.soundex_boost}]'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladmir Pootin\"}' | jq -c '{alert_reason, score: ([.matches[].score] | max)}'"],"expect":"step 2 shows the live path raising an alert below the threshold with alert_reason \"name_near_miss\"; step 1 shows monitoring alerts reaching the same decision — any raised on proximity carry the same alert_reason, and no alert reports a soundex_boost of zero, since that field is written only when phonetic agreement contributed","audit":"alert_reason and the phonetic contribution are stored with the alert, not only returned, so a reviewer asking why a below-threshold monitoring alert exists reads the answer off the record","regression":"TestReverseScreenNearMissAlerts, TestReverseScreenAlertCarriesBoost, TestMonitoringPipelineMatchesForwardOnSharedSteps, TestMonitoringPipelineUsesTheRequestName, TestReverseScreenScale"}},{"date":"2026-09-15","category":"added","summary":"The changelog can now be watched: it carries an ETag, an entry count and a content fingerprint, so a conditional request tells you whether anything changed without downloading and comparing every entry","details":"If you run a control that watches this vendor for behaviour changes, it previously had nothing stable to watch. Entries carry no identifier, several can share a date, and the version header deliberately does not move for additive or corrective changes — so the only correct way to detect a change was to store every entry and compare them on each poll. A customer operating exactly that control missed four behaviour changes we shipped in one day, and concluded the fault was theirs. It was ours. The response now carries entry_count and fingerprint, and the same fingerprint is served as an ETag: send it back as If-None-Match and an unchanged changelog answers 304 with no body. The fingerprint covers the entire content of every entry, so nothing can change without it changing. WHAT YOU WILL SEE: two new fields and a validator header. Polling with If-None-Match costs one request and tells you plainly whether to look. Additive — no existing field changed meaning and no version is minted.","endpoint":"/v1/changelog","verification":{"steps":["curl -s \"$BASE/v1/changelog\" | jq -c '{entry_count, fingerprint, actual: (.entries | length)}'","ETAG=$(curl -sI \"$BASE/v1/changelog\" | awk 'tolower($1)==\"etag:\"{print $2}' | tr -d '\\r'); curl -s -o /dev/null -w '%{http_code}\\n' -H \"If-None-Match: $ETAG\" \"$BASE/v1/changelog\""],"expect":"step 1 returns an entry_count equal to the number of entries and a non-empty fingerprint that is identical across repeated requests; step 2 returns 304, and a request with no If-None-Match still returns the full list with 200","audit":"the fingerprint in force on a given date can be recorded alongside your own change-control evidence, so what this vendor was publishing at that moment is provable afterwards rather than asserted","regression":"TestChangelogIsDetectable, TestChangelogFingerprintCoversEveryField"}},{"date":"2026-09-14","category":"fixed","summary":"Date of birth is compared on the Australian list, where it was silently ignored; a record that lists several possible birth dates is now compared against all of them; an attribute that could not be evaluated always says so; every alert names the rule that raised it; and a match one character from part of a longer name no longer alerts when the rest of that name is absent","details":"Five changes, all of them about the same thing: a check that cannot run must say so rather than look like a check that passed. FIRST, dates of birth on the Australian list were being ignored. That publisher ships its list as a spreadsheet, and a date in a spreadsheet is a number; the value was stored in that raw form, nothing downstream could read it as a date, and the comparison quietly did not happen. A correct date of birth, a date of birth decades wrong, and no date of birth at all produced identical results on that list while every other list discriminated normally. Dates are now decoded when the list is loaded, and existing records were repaired. SECOND, publishers often record several possible birth dates for one person, sometimes with a note that they are approximate. Those values could not be read either. A record that lists several dates is now compared against all of them, and only contradicts you if your date matches none — matching any one of the dates the publisher itself offers is agreement, not a conflict. THIRD, when an attribute cannot be evaluated at all, the response now says so instead of staying silent. Before, an unreadable value was indistinguishable from a value that agreed. FOURTH, a match that nothing else could explain now carries a signal saying no attributes were supplied to discriminate with, so a full-confidence alert no longer arrives with an empty explanation and send us a date of birth or a country is stated rather than implied. FIFTH, the alert raised for a name a character or two from a listed name now requires that the part of the listed name identifying the person be present in your request at all. It was enough for any single part of a longer record to be close, so a given name plus a patronymic could alert against someone whose surname appeared nowhere in the request. AND: every alert now carries alert_reason naming which rule raised it, on the screening response and on the stored alert. WHAT YOU WILL SEE: on the Australian list, supplying a date of birth now changes the outcome where it previously did not, and a wrong one lowers the score as it does elsewhere. Some responses carry additional zero-impact signals reporting checks that could not run — these never move a score. Slightly fewer alerts of the near-miss kind. Read alert_reason rather than inferring the rule from the score; it is absent on alerts created before today, which were raised before the reason was recorded. Additive — no existing field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"lists\":[\"australia\"]}' | jq -r '.matches | max_by(.score) | .date_of_birth'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"date_of_birth\":\"1990-01-01\",\"lists\":[\"australia\"]}' | jq -c '{score: (.matches|max_by(.score)|.score), signals: [(.matches|max_by(.score)).dismissal_signals[] | {reason_code, impact_on_score}]}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladmir Pootin\"}' | jq -c '{alert: (.alert_id != null), alert_reason}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Sergei Ivanov\"}' | jq -c '.matches | max_by(.score) | {score, signals: [.dismissal_signals[].reason_code]}'"],"expect":"step 1 returns a real date, never a bare integer; step 2 returns a score well below the one the same request gets without a date of birth, carrying a signal naming the date-of-birth conflict with a negative impact — and supplying no date of birth at all still matches and still alerts, because this applies the attribute when given rather than requiring it; step 3 returns an alert with alert_reason \"name_near_miss\", a rule that raises an alert below your threshold and now says so by name; step 4 returns a top match that carries at least one signal — the alert is correct and is not suppressed, but it no longer arrives with nothing attached","audit":"alert_reason is stored with the alert, not only returned, so a reviewer asking why a below-threshold match raised an alert reads the answer off the record rather than re-deriving it; alerts created before this change report it as absent rather than inferring one","regression":"TestNormalizeDOBDecodesExcelSerials, TestNormalizeDOBLeavesEverythingElseAlone, TestImportFixtureDecodesDOBAndIsStable, TestCompareDOBValueMultiValue, TestCompareDOBValueKeepsSingleValueListsUnchanged, TestNearMissRequiresIdentifyingComponent, TestApplyInsufficientDataSignal, TestAlertReasonNamesTheRule, TestSeededMissBenchmark, TestFalsePositiveBenchmark"}},{"date":"2026-09-14","category":"fixed","summary":"A name where one part is truncated or a character off now reaches the designation it belongs to, instead of being found and then set aside; matches are ordered highest score first; and a near miss raises an alert on batch screening as it already did on a single screen","details":"Three changes that move in the same direction: a match that should reach a reviewer now does. FIRST, the control that sets aside a hit when only one part of a longer name matched — OFAC's own guidance that a surname alone is not a valid match — was applying when you had in fact supplied every part of the name and one of them was damaged. Screening a person whose first name was cut short, or carried a single letter of transliteration difference, found the listed party and then set the result aside as though you had supplied nothing for that name. It no longer does: a recognisably degraded form of a name counts as having been supplied. A hit corroborated by only one part of a name is still set aside, exactly as before. SECOND, matches[] is now ordered by the score the response reports, highest first, so the first element is the top match. It was ordered before the response's own adjustments were applied and never reordered, so the first element could be a low-scoring match while a higher one sat further down; the summary and the alert decision always used the highest, so the response disagreed with itself rather than being wrong. Equal scores keep a stable order across identical requests. THIRD, a match sitting one or two characters from a listed name raises an alert on batch screening, which it already did on a single screen. WHAT YOU WILL SEE: some names that previously returned a low score now return a high one and may raise an alert — these are matches the engine had already found. On a 3,375-name benchmark of ordinary customer names the change added two alerts. If you read matches[0], you now get the top match; if you scan for the maximum, nothing changes. Batch results may carry alerts they did not before, each explaining itself through the same signal the single-screen path uses. No field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Hass Aweys\",\"lists\":[\"eu\"]}' | jq -c '{alert: (.alert_id != null), top: ([.matches[]] | max_by(.score) | {uid, score, signals: [.dismissal_signals[].reason_code]})}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Wei Chen\",\"country\":\"CN\"}' | jq -c '[.matches[]] | max_by(.score) | {score, signals: [.dismissal_signals[].reason_code]}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Fatima Al-Sayed\"}' | jq -e '[.matches[].score] as $s | ($s == ($s | sort | reverse)) and (.matches[0].score == ($s | max))'"],"expect":"step 1 returns the listed designation at a score above 90 with an alert, and its signals do NOT include partial_name_match_single_token — the truncated first name no longer sets the match aside; step 2 still carries partial_name_match_single_token, so a genuinely one-part match is still discounted; step 3 exits 0, meaning matches[] is ordered descending and matches[0] is the highest-scoring match","audit":"the ordering applies to the matches persisted with the screening as well as the ones returned, so an exported decision and the response a reviewer saw list them identically","regression":"TestApplyNameTokenSignalSparesDegradedTokens, TestDegradedFormRejectsDifferentNames, TestApplyNameTokenSignalFiresOnFoldCollision, TestFinalizeScoresOrdersByFinalScore, TestBatchNearMissAlerts, TestSeededMissBenchmark, TestFalsePositiveBenchmark"}},{"date":"2026-09-14","category":"added","summary":"A match one or two characters from a listed name now raises an alert and says why, where it previously returned a clean-looking weak result with nothing attached","details":"A one-character difference from a published sanctions name is a fundamentally different object from an unrelated name that happens to score similarly, and until now the response could not tell them apart. Screening a listed name with a single letter changed could return a weak-match score, raise no alert, and carry no explanation at all — the score was plain literal similarity and the threshold fell in the gap. Responses now carry a name_near_miss entry in dismissal_signals when a match sits within a very small edit distance of the listed name AND is already close to the alert threshold. It reports the two tokens that nearly matched and the number of characters between them, so a reviewer or an agent can act on the proximity rather than inferring it from a score. Both conditions are required, and the second is not decoration: character proximity alone does not distinguish a deliberate near-miss from an ordinary coincidence, because an unrelated name can sit exactly as close lexically as a real evasion does — we tested that directly and the two are inseparable on distance alone. The signal never changes a score or a verdict. It does, however, raise an alert: a match flagged this way alerts even though its score sits below your threshold, because a one-character difference from a published sanctions name is the shape a deliberate evasion takes, and under strict liability the expensive direction is the miss. WHAT YOU WILL SEE: a small number of additional alerts, each carrying the signal that explains why it was raised despite the score, and each identifying the two tokens and the characters between them so it can be dispositioned quickly. Across a 3,375-name benchmark of ordinary customer names the signal appears on eighteen — well under one percent — and those are the only additional alerts it produces. Scores themselves are untouched, so any threshold you have tuned against Noble scores continues to behave as before. Additive — no existing field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Rafael Caro Quintebo\"}' | jq -c '{alert: (.alert_id != null), top: ([.matches[]] | max_by(.score) | {score, near_miss: [.dismissal_signals[] | select(.reason_code==\"name_near_miss\") | {request_value, target_value, edit_distance, impact_on_score}]})}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Rafael Caro Quintero\"}' | jq -c '{score: ([.matches[].score] | max), alert: (.alert_id != null)}'"],"expect":"step 1 returns a match scoring under the alert threshold that nonetheless carries a non-null alert_id, plus a name_near_miss signal whose request_value and target_value differ by one character, with edit_distance 1 and impact_on_score 0 — the score is untouched, the proximity is stated, and the alert is raised on the proximity rather than the score; step 2, the name spelled correctly, returns 100 with an alert as it always has","audit":"the proximity is recorded on the persisted screening, so a reviewer of a below-threshold result can see it was a character from a listed name rather than having to re-derive that from the score","regression":"TestApplyNameNearMissSignal, TestLevenshtein, TestNearMissAlertsBelowThreshold, TestFalsePositiveBenchmark"}},{"date":"2026-09-14","category":"fixed","summary":"Screening an alias by its complete published name now matches it in full — previously only part of the name was searchable, so an exact alias could return a weak score and no alert","details":"OFAC publishes alternate names for designated parties, and stores many of them as separate given and family name fields. Noble searches those aliases three ways: the given name, the family name, and the two joined together. The joined form was not being maintained — it was filled in once when the column was introduced and never updated afterwards, so for aliases loaded since then it was empty and the search index built on it held nothing. The practical effect was that supplying an alias exactly as OFAC publishes it could only ever be compared against one half of it. One published alias of a designated party returned a score of 77.8 and no alert, where matching the whole string returns 100 and alerts. Across the OFAC alias set, 8,622 aliases that carry both a given and a family name could not reach the alert threshold on an exact match of their own published name; every one of them now does. The joined form is now derived by the database itself rather than maintained by the import, so it cannot fall out of step with the data again, and a test now fails the build if any search column in any sanctions list is left to be maintained by hand. WHAT YOU WILL SEE: more matches and some higher scores where an alias is involved, and alerts on exact alias matches that previously returned a weak result. This restores matching that was always intended; it is not a loosening of the threshold. Measured against the 3,375-name benchmark corpus the change adds eight alerts in total, so the recall is recovered at very little cost in noise. Request and response shapes are unchanged.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"MAKLED GARCIA, Valed\"}' | jq -c '{score: ([.matches[].score] | max), alert: (.alert_id != null), uid: ([.matches[]] | max_by(.score) | .uid)}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"MAKLED GARCIA, Walid\"}' | jq -c '{score: ([.matches[].score] | max), alert: (.alert_id != null)}'"],"expect":"step 1 returns score 100 with an alert against uid 11500 — \"MAKLED GARCIA, Valed\" is an alias OFAC publishes for that designation, and it previously returned 77.8 with no alert; step 2, the primary name, returns 100 with an alert as it always has","audit":"an exact match on a published alias is now recorded at full strength with the alert it warrants, so a reviewer of the screening record sees the designation was hit rather than a weak score with no alert attached","regression":"TestSearchColumnsAreGenerated, TestOFACAliasSearchColumnPopulated, TestFalsePositiveBenchmark, TestScreenAliasQuality"}},{"date":"2026-09-13","category":"changed","summary":"Phonetic name agreement now scales with how much of the name literally matched, so a common-surname collision no longer reaches the alert threshold on phonetics alone","details":"A match score is literal name similarity plus a phonetic agreement amount, and that amount was flat: a surname agreeing phonetically contributed the same fixed number of points whether the literal similarity was high or very low. On a common surname the phonetic component was therefore most of the score while the literal match contributed almost nothing. Measured against production, five of six alerts raised by a twenty-name battery of ordinary customer names were carried over the threshold this way — one reached a score of 86.3 against a listed party sharing only a surname, with an entirely different given name. The phonetic amount is now weighted by the literal similarity it corroborates, so the same agreement contributes more to a strong name match than to a weak one. Phonetic agreement is corroborating evidence, and it should amplify a real match rather than manufacture one. WHAT YOU WILL SEE: fewer alerts, and lower scores on matches that were resting on phonetic agreement rather than on the name itself. Recall was not traded for it — on the benchmark that measures MISSED listed parties, five of seven sanctions lists are unchanged and two move by a single variant each. A more aggressive weighting was measured and deliberately rejected because it DID trade recall: a transliteration variant has low literal similarity by construction and is exactly the case phonetic matching exists to catch, so weighting too hard on literal similarity penalises it. Scores are directly comparable before and after only for matches with no phonetic component; if you have tuned thresholds of your own against Noble scores, re-check them. Request and response shapes are unchanged.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Fatima Al-Sayed\"}' | jq -c '[.matches[]] | max_by(.score) | {score, target: (.first_name + \" \" + .last_name), boost: [.dismissal_signals[] | select(.reason_code==\"phonetic_name_boost\") | .impact_on_score]}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Ahmed Hassan\"}' | jq -c '[.matches[]] | max_by(.score) | {score, target: (.first_name + \" \" + .last_name)}'"],"expect":"in step 1 the phonetic contribution reported in impact_on_score is strictly less than it would be for a fixed amount, and strictly less than the same agreement contributes to step 2's stronger literal match — the contribution scales with literal similarity, which is the guarantee this change makes. Step 2 still returns 100 against the listed \"Ahmad Hassan\", the transliteration case phonetic matching exists for, showing recall intact. (Step 1 also falls below the alert threshold today, but read the scaling relation rather than that margin: it is narrow, and an ordinary list update can move it either way without anything changing here.)","audit":"every score that includes a phonetic component reports that component's actual contribution in dismissal_signals, so an analyst reviewing an alert can see how much of the score came from literal similarity and how much from phonetic agreement","regression":"TestApplyPhoneticNameBoostSignal, TestFalsePositiveBenchmark, TestSeededMissBenchmark, TestReverseScreenAIVariationParity, TestScreenBatchVariantParity"}},{"date":"2026-09-13","category":"fixed","summary":"An exact match on an alias OFAC publishes is no longer halved as a partial-name match — it scores in full and raises an alert","details":"OFAC FAQ 5 Step 3 asks whether just one of two or more names matches, i.e. just the last name, and directs that such a hit is not a valid match. Noble applies that as a scoring discount. The discount was being measured against the designation's PRIMARY name even when the match came from an alias — and an alias, by construction, shares few or no tokens with the primary name. So screening the exact alias string OFAC publishes for a designated party matched at full similarity and was then halved for overlapping the primary name in only one token, landing under the alert threshold. An exact hit on a published designation raised nothing. The discount is now skipped when any pathway matched a multi-token name the publisher lists — a primary name, an alias, or a stored variation — in its entirety, because in that case all of the name matched and the rule's premise does not hold. The control itself is unchanged in every other case: a bare surname against a multi-token designation still takes the discount, including when that surname is itself a published alias, and an approximate alias hit still takes it. Alias strength is unaffected — a weak alias is still reported as weak, it is simply no longer discounted twice. Measured against the 3,375-name benchmark corpus the change moves the alert rate not at all (1094/3375 before and after); it is narrow by construction and affects only exact matches on multi-token published names.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"EL TURCO WALID\"}' | jq -c '{score: .matches[0].score, uid: .matches[0].uid, alert: .alert_id, signals: [.matches[0].dismissal_signals[].reason_code]}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"WALID\"}' | jq -c '{score: .matches[0].score, signals: [.matches[0].dismissal_signals[].reason_code]}'"],"expect":"step 1 returns the uid-11500 match at score 100 with an alert_id and no partial_name_match_single_token signal — \"EL TURCO WALID\" is an alias OFAC publishes for that designation; step 2, a bare surname, still carries partial_name_match_single_token and scores 50, showing the FAQ 5 Step 3 control intact","audit":"a screening whose input reproduces a published alias in full now records a full-strength match and the alert it warrants, so the audit trail shows the designation was hit rather than showing a below-threshold score with no alert","regression":"TestApplyNameTokenSignalSparesFullPublishedAlias, TestApplyNameTokenSignal, TestApplyNameTokenSignalSpares150Variants, TestFalsePositiveBenchmark, TestSeededMissBenchmark"}},{"date":"2026-09-13","category":"changed","summary":"The CSV export gains a jurisdiction_nexus_status column, so a flat export can tell a geography we hold no determination about from one we evaluated and cleared","details":"The CSV export carried the declared region and the embargo advisory, which meant a screening whose declared region we hold no determination about and one whose region we evaluated and excluded rendered identically: a region beside a blank advisory. The structured responses distinguish these, and CSV is the format pulled for a regulatory review precisely because it is flat and readable, so it was the wrong surface to leave ambiguous. A jurisdiction_nexus_status column is added carrying embargoed, evaluated_not_embargoed, or no_determination, and left blank when the screening declared no geography at all — blank means nothing was declared, not nothing was found. The column is inserted after jurisdiction_authority and before deep_screen. A consumer reading columns by NAME is unaffected. A consumer reading by fixed position must re-read the header row: this is the second such change in ten days, after the three columns added on 2026-09-03, and both were made before the export had positional consumers we know of. The JSON export is unchanged and remains the recommended integration.","endpoint":"/v1/export","verification":{"steps":["T1=$(curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"UA\",\"region\":\"Crimea Republic\"}}' | jq -r .trace_id)","T2=$(curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"UA\",\"region\":\"Kherson\"}}' | jq -r .trace_id)","curl -s \"https://api.noblesight.io/v1/export?format=csv\u0026trace_id=$T1\" -H \"X-API-Key: $KEY\" | head -1 | tr ',' '\\n' | grep -n jurisdiction_nexus_status","for T in $T1 $T2; do curl -s \"https://api.noblesight.io/v1/export?format=csv\u0026trace_id=$T\" -H \"X-API-Key: $KEY\" | tail -1 | awk -F, '{print $14}'; done"],"expect":"step 3 reports jurisdiction_nexus_status at column 14; step 4 prints \"no_determination\" then \"evaluated_not_embargoed\" — two rows that were previously identical","audit":"a CSV pull is now sufficient to establish whether a declared geography was one Noble holds a determination about, so a reviewer can tell a disclosure that did not land from one that was evaluated and cleared without returning to the API","regression":"TestExportCSVRowNexusStatus, TestExportCSVRowJurisdiction, TestExportCSVRow"}},{"date":"2026-09-13","category":"fixed","summary":"Replaying a point-in-time screening no longer reports a false mismatch on its jurisdiction advisory","details":"A screening submitted with as_of asks what was legally true on a past date, and is issued the advisory that applied then. Replay was re-deriving that advisory from the record of what our embargo table has carried over time, resolved at the same past date — and that record begins when the advisory feature shipped. A screening as of any date before then therefore replayed with no advisory at all, contradicted the record it was verifying, and returned a verdict of mismatch for a screening that had in fact reproduced exactly. On the endpoint whose purpose is demonstrating that stored decisions reproduce, a false alarm is the most costly kind of wrong. Replay now keeps the two questions apart: the time the screening ran determines which version of the embargo table applied, and the as-of instant determines what was legally in force. Both were previously collapsed onto one. Ordinary screenings without as_of are unaffected, and the behaviour that motivated the original design is preserved — a screening issued under a program later revoked still replays carrying the advisory it was given, and a screening from before the advisory existed still replays with none.","endpoint":"/v1/export/replay","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"country\":\"IR\",\"as_of\":\"2024-01-01\"}' | jq -c '{trace: .trace_id, j: .jurisdiction_risk.jurisdiction}'","curl -s -X POST https://api.noblesight.io/v1/export/replay -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"trace_id\":\"\u003ctrace_id from step 1\u003e\"}' | jq -c '{verdict, j: .replayed.jurisdiction_risk.jurisdiction, diffs: [.differences[].field]}'"],"expect":"step 1 returns jurisdiction \"IR\"; step 2 returns verdict \"match\" with the same jurisdiction \"IR\" and a differences array containing no jurisdiction field. Before this fix step 2 returned verdict \"mismatch\" with jurisdiction_risk.jurisdiction listed as stored \"IR\", replayed \"absent\".","audit":"a point-in-time screening and its replay now agree on the embargo that applied, so the reproduction evidence an examiner is shown does not contradict the record it verifies","regression":"TestReplayOfAsOfScreeningMatches, TestReplayOfAsOfScreeningReproducesItsAdvisory, TestReplayStillHonoursTheKnowledgeTimeline"}},{"date":"2026-09-13","category":"added","summary":"Screening responses now report which geographic declaration was evaluated for an embargo nexus, so a region we have no determination about is no longer indistinguishable from one we evaluated and cleared","details":"The embargo advisory answers what was found. Nothing answered what was looked at, and on a signal that only fires for geography you declare, that gap mattered: a request declaring the region \"Kherson\" — evaluated, with a cited determination that no covered-region designation exists — and a request declaring a value we hold no determination about both returned the same empty answer. A caller who mistyped a region believed they had disclosed a nexus and had not, and nothing in the response said otherwise. A new jurisdiction_nexus object reports the declaration that was evaluated, the request field it came from, and one of three statuses: embargoed (always accompanied by the advisory), evaluated_not_embargoed (we hold a determination and it is not an embargo), or no_determination (we hold none for what you sent). It is present whenever a request declares any geography and absent when none is declared. Because precedence is first-resolving, the echoed value is not always the most specific field supplied: declaring country IR with region IR-07 reports IR, since we hold no determination about that province and the country embargo therefore stands — the fall-through is now visible rather than inferred. The object is persisted with the screening, returned per item on batch, and reproduced on replay. Additive: no existing field changed meaning and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"UA\",\"region\":\"Crimea Republic\"}}' | jq -c '{risk: .jurisdiction_risk, nexus: .jurisdiction_nexus}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"UA\",\"region\":\"Kherson\"}}' | jq -c '{risk: .jurisdiction_risk, nexus: .jurisdiction_nexus}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"IR\",\"region\":\"IR-07\"}}' | jq -c '{risk: .jurisdiction_risk.jurisdiction, nexus: .jurisdiction_nexus}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\"}' | jq -c '.jurisdiction_nexus'"],"expect":"step 1 returns a null risk with nexus status \"no_determination\", evaluated \"Crimea Republic\", source \"location.region\"; step 2 returns a null risk with status \"evaluated_not_embargoed\" — the two are no longer the same answer; step 3 returns jurisdiction \"IR\" with nexus evaluated \"IR\" and source \"location.country\", showing the province deferred to its country; step 4 returns null, because no geography was declared","audit":"the persisted screening now records the interpretation alongside the declaration, so a reviewer can establish which geographic input the engine acted on rather than inferring it from the precedence rules","regression":"TestScreenJurisdictionNexusDisclosure, TestScreenNexusPersisted, TestJurisdictionNexusStatuses, TestJurisdictionNexusAgreesWithAdvisory, TestNexusEmbargoBeatsExclusion, TestNexusOnRecordFollowsTheTableNotToday, TestScreenDecodesJurisdictionNexus, TestBatchDecodesJurisdictionNexus, TestCompareReplayNexusAsymmetric, TestSummarizeScreenNarratesUnrecognizedNexus"}},{"date":"2026-09-13","category":"added","summary":"A match that nothing reduced now reports what raised its score, so an alert is no longer the one result in a screening with no explanation attached","details":"Each match carries structured signals explaining its score, but until now those recorded only reductions. A reduction large enough to record usually drops a match below the alert threshold, so the matches that survive to raise an alert were the ones nothing had reduced — and they arrived with an empty explanation while lower-scoring matches in the same response were fully explained. An analyst opened the alert, saw a high score with nothing attached, and had to investigate from scratch. A match score is its literal name similarity plus a phonetic name agreement, and that phonetic amount is flat: it does not scale with how much of the name matched, so on a common surname it can be most of the score while literal similarity stays low. Where nothing reduced a match, the response now carries a phonetic_name_boost entry naming how many points came from phonetic agreement rather than literal similarity, and the list text that produced it. It is the one entry whose score impact is positive, and it never changes a score — it accounts for one. Matches that were already reduced are unchanged, so the reason an analyst needs is not buried under a second entry. Each match pathway additionally reports soundex_boost alongside its existing trigram_score, so the two can be read as the sum they are. Additive: fields were added and none changed meaning, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Maria Garcia\",\"country\":\"MX\"}' | jq -c '.matches | max_by(.score) | {score, signals: [.dismissal_signals[] | {reason_code, impact_on_score}], sources: [.match_sources[] | {trigram_score, soundex_boost}]}'","Assert the invariant on any screening that raised an alert: the top-scoring match carries at least one signal."],"expect":"the top match carries at least one dismissal_signals entry; where nothing reduced it that entry is reason_code \"phonetic_name_boost\" with a positive impact_on_score, and its match_sources show a trigram_score and a soundex_boost that sum to the pathway's contribution","audit":"the signal is recorded on the persisted screening response, so GET /v1/export?trace_id= shows for any screening in the retention window how much of a match's score came from phonetic agreement rather than literal similarity","regression":"TestApplyPhoneticNameBoostSignal covers the explained, already-penalized, and no-boost cases and asserts the score is never changed"}},{"date":"2026-09-12","category":"fixed","summary":"A name matching a listed person on only one short token is now discounted and says so — previously some such matches scored as though the whole name agreed, and reached the alert threshold with no explanation attached","details":"Matching a person on just one token of a multi-token name is not a valid match on its own, so those matches are discounted and carry a partial_name_match_single_token reason code. Deciding which tokens agree uses phonetic equivalence, so the discount does not misfire on transliteration and spelling variants the engine correctly treats as the same name. That equivalence was too permissive in one respect: it generalized a name's first letter to a class of eight letters, which made some unrelated short surnames compare as equal — so a query agreeing on a first name only could read as agreeing on the whole name, take no discount, and arrive as an alert with an empty explanation. Token agreement now additionally requires the phonetic code to carry enough detail to be meaningful once the first letter has been generalized, which it does for genuine variants and does not for unrelated short names. Affected matches score lower and now carry the reason code, so some that previously raised alerts no longer will. Transliteration variants are unchanged and still take no discount — that direction is guarded by tests in both directions, because discounting a real variant would push a confirmed match below the alert threshold. No request or response shape changed.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Wei Chen\",\"country\":\"CN\"}' | jq -c '.matches | max_by(.score) | {score, matched: [.match_sources[].matched_text], signals: [.dismissal_signals[].reason_code]}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Kim Chong Un\"}' | jq -c '.matches | max_by(.score) | {score, signals: [.dismissal_signals[].reason_code]}'"],"expect":"the first call's top match either no longer agrees on a single short token or carries reason_code \"partial_name_match_single_token\" with a negative impact_on_score; the second call still returns a top score of 100 with no partial_name_match_single_token entry, because a first-letter transliteration variant agrees on the whole name","audit":"the discount and its reason code are recorded on the persisted screening response, so GET /v1/export?trace_id= shows which tokens were held to agree and what the discount cost, for any screening in the retention window","regression":"TestSoundexNameTokenMatch and TestSoundexNameTokenMatchStricterThanFold lock the equivalence in both directions; TestApplyNameTokenSignalFiresOnFoldCollision covers the reported case; TestApplyNameTokenSignalSpares150Variants proves transliteration variants still take no discount"}},{"date":"2026-09-12","category":"fixed","summary":"Match scores and score impacts are reported to one decimal place instead of carrying meaningless trailing precision — a score that read 91.15384638309479 now reads 91.2","details":"Name similarity is computed in the database at single precision and widened for scoring, so every score carried a fraction of representation residue, and multiplying by attribute penalties compounded it. The result reached the API raw: scores rendered as 91.15384638309479 and 24.500000208616257, an average_score of 52.36538474261761, and score impacts of -21.000000178813934. Those numbers are read by compliance analysts and quoted in examination evidence, where fifteen digits imply a confidence the matching engine does not have and invite a reviewer to reconcile figures that were never meant to be exact. Scores, the mean score, and each dismissal signal's score impact are now rounded to one decimal place, which is the engine's real resolution. Rounding is applied where scores are calculated rather than when they are serialized, so the score returned, the score stored for the audit record, and the score compared against the alert threshold are the same number — a response that disagreed with its own audit record would be the more serious defect. One consequence worth stating: at an alert threshold this can only round a score up into the range that raises an alert, never out of it, so the change can add an alert for a score within a tenth of the threshold and can never suppress one. No request or response shape changed and no field was added or removed.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"country\":\"US\"}' | jq -c '{average_score, scores: [.matches[].score], impacts: [.matches[].dismissal_signals[].impact_on_score]}'","Assert that every number printed has at most one digit after the decimal point."],"expect":"average_score, every match score, and every impact_on_score carry at most one decimal place — for example 91.2 rather than 91.15384638309479, and -21 rather than -21.000000178813934","audit":"the rounded score is the value persisted with the screening record and the value compared against the alert threshold, so GET /v1/export?trace_id= returns the same figures the response carried and a reviewer reconciling an alert against its screening finds no discrepancy","regression":"TestRoundScore locks the reported precision against the observed float artifacts; TestRoundScoreThresholdDirectionIsFailSafe proves rounding can never move a score out of alert-firing range"}},{"date":"2026-09-12","category":"fixed","summary":"A match is no longer silently left unscored on an attribute you supplied — when the matched record carries no country or date of birth to compare against, the response now says so instead of saying nothing","details":"Screening compares a supplied country or date of birth against the matched record and reduces the score when they contradict, recording a dismissal_signals entry with the reason and the exact points deducted. When the matched record carried no value for that attribute, no comparison was possible and nothing was recorded — so a match that was never compared looked identical to one that was compared and agreed: full score, empty dismissal_signals. Attribute coverage is uneven across sources, so this was not rare, and it let an unpenalized match from a list carrying no countries outrank a penalized match for the same person with nothing in the response marking the difference. Such a match now carries a dismissal_signals entry with reason code country_not_comparable or dob_not_comparable, severity info, and impact_on_score 0. These record an evaluation outcome, not dismissal evidence: they report that the comparison could not be made, and must not be read as weak evidence in the match's favour. No score changes as a result of this, and the existing mismatch reason codes are unaffected. severity is a new value on an existing field and impact_on_score is always 0, so this is additive and no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"country\":\"US\"}' | jq -c '[.matches[] | {source, score, signals: [.dismissal_signals[] | {reason_code, severity, impact_on_score}]}] | .[0:4]'","Assert the invariant across every match in that response: no match is missing a country signal. Each one either carries country_mismatch with a negative impact_on_score, or country_not_comparable with impact_on_score 0."],"expect":"every match carries a country dismissal_signals entry: records holding a contradicting country show reason_code \"country_mismatch\" with a negative impact_on_score, and records holding no country show reason_code \"country_not_comparable\" with severity \"info\" and impact_on_score 0. No match carries an empty dismissal_signals array for an attribute that was supplied","audit":"the persisted screening response (GET /v1/export?trace_id=) preserves every signal including the not-comparable ones, so a reviewer can establish from the stored record alone whether an attribute was evaluated and agreed or could not be evaluated at all","regression":"TestNotComparableSignal locks the shape and the 1.0 multiplier; TestApplyPenaltyNotComparableLeavesScore proves the score is untouched and legacy and structured reasons stay in lockstep"}},{"date":"2026-09-10","category":"fixed","summary":"Canada (SEMA) screening is current again — the publisher renamed every field in its feed, which stopped imports silently on 2026-08-24 while screening continued against the last good snapshot","details":"Global Affairs Canada republished the SEMA consolidated list with bilingual field names. Noble read the feed by the previous names, so every record arrived with an empty identity. The import refused the file rather than loading 5,698 unidentifiable records — the correct outcome, and screening continued uninterrupted against the 2026-08-19 snapshot, so no screening ran against corrupt data. What was wrong is that the refusal was quiet: GET /v1/lists reported canada as stale and GET /status reported degraded, exactly as designed, but neither distinguished a publisher format change from any other cause, and the condition persisted for seventeen days. Canada now imports on the normal polling cadence, and any future rename of a field fails the build rather than reaching production. Customers screening against canada should expect designations and delistings published after 2026-08-19 to appear. No request or response shape changed, so no version is minted.","endpoint":"/v1/lists","verification":{"steps":["curl -s https://api.noblesight.io/v1/lists | jq -c '.lists[] | select(.list==\"canada\") | {status, publish_date, record_count, last_checked, last_import}'","curl -s https://api.noblesight.io/status | jq -c '{status, lists: .components.lists.status}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Alexander Lukashenko\",\"lists\":[\"canada\"]}' | jq -c '{screened: .lists_screened, versions: .list_versions, count: .match_count}'"],"expect":"canada reports status \"ok\" with last_checked inside the last polling interval and a publish_date on or after 2026-09-10; /status reports \"operational\" with components.lists \"ok\"; and a canada-scoped screening returns list_versions naming canada at the new publish id","audit":"each import is recorded with its publisher snapshot id and record count, and every screening persists the list_versions it ran against, so a Canada screening from any date can be replayed against the exact snapshot that produced it","regression":"TestParseXMLBindsEveryElement asserts every SEMA field reaches its parsed value and fails on a rename; TestParseXMLLiveFixtureVocabulary checks the published feed still uses those names; TestParseXMLRejectsDuplicateSids locks the fail-closed guard that refused the unreadable file"}},{"date":"2026-09-10","category":"fixed","summary":"The public site now reports each sanctions list's real ingestion status instead of a fixed label, and the list endpoint is named correctly for AI agents","details":"The coverage grid on the public site labelled every list \"Live\" as static text, so it made a freshness claim independent of the API and could not follow a list into a stale state. Each badge is now resolved from GET /v1/lists when the page loads and reads Live or Stale to match, with the last import and last check times available on hover. With scripting unavailable the badges assert nothing rather than defaulting to Live, and the grid links to /v1/lists and /status so the claim is checkable at its source. A stale list is still screened, against its last imported snapshot, and /status continues to report degraded while any list is stale — that behaviour is unchanged; only the public claim now matches it. Separately, the machine-readable guide at /llms.txt directed agents to GET /v1/sanctions-lists, which is not a route and returns 404; it now names GET /v1/lists and points at the per-list status and last-checked fields as the authority on freshness. No API request or response shape changed.","endpoint":"/v1/lists","verification":{"steps":["curl -s https://api.noblesight.io/v1/lists | jq '[.lists[] | {list, status}]'","curl -s -o /dev/null -w '%{http_code}\\n' https://api.noblesight.io/v1/sanctions-lists   # 404 — the name corrected in /llms.txt","Open https://noblesight.io/#coverage and compare each card's badge against step 1."],"expect":"every card reads Live for a list whose status is \"ok\" and Stale for any list whose status is not \"ok\", matching step 1 exactly; step 2 returns 404, and /llms.txt no longer cites that path","audit":"the page reads the same public, unauthenticated GET /v1/lists that step 1 calls, so the badge and the API cannot disagree; each badge also exposes the list's last_import and last_checked on hover for comparison against the same fields in the response","regression":"TestMarketingBadgesDoNotHardcodeLive, TestLLMsTxtNamesRealEndpoints"}},{"date":"2026-09-09","category":"fixed","summary":"A service-side failure is no longer reported as an authentication, permission, or not-found error — those now mean what they say, and a transient fault returns 5xx so it can be retried","details":"Several endpoints treated any failure during a lookup as a statement about the request: a key that could not be verified was reported as an invalid API key, an account whose tenant could not be read was reported as not provisioned, and a record that could not be fetched was reported as not found. All three are indistinguishable from the real thing, and all three invite a damaging response — rotating a credential that was never bad, opening a support ticket about an account that is fine, re-enrolling a monitored name that is already enrolled, or abandoning a batch job that is still running. Fault attribution is now a contract across the API: a 4xx describes the request or the account and will still be true on retry, while a service-side failure always returns 5xx with type api_error. A genuinely unknown key still returns 401, an account without a tenant still returns 403, and a record that genuinely does not exist still returns 404 — the change narrows those answers to the cases where they are true rather than removing them. One further case: a batch whose results could not be read now returns an error instead of a completed job with an empty result list. Clients may treat 4xx as terminal and 5xx as retryable. No request or response shape changed.","endpoint":"/v1/export","verification":{"steps":["curl -s -o /dev/null -w '%{http_code}\\n' -X GET 'https://api.noblesight.io/v1/export?trace_id=definitely-not-a-real-trace-id' -H \"X-API-Key: $KEY\"","curl -s -X GET 'https://api.noblesight.io/v1/export?trace_id=definitely-not-a-real-trace-id' -H \"X-API-Key: $KEY\" | jq -c '{type, code}'","curl -s -o /dev/null -w '%{http_code}\\n' -X GET https://api.noblesight.io/v1/webhooks -H 'X-API-Key: noble_live_0000000000000000000000000000000000'"],"expect":"the first two calls return 404 with {\"type\":\"not_found_error\",\"code\":\"not_found\"}, which now means the record is genuinely absent rather than that the lookup failed; the third returns 401 with code \"api_key_invalid\". Any 4xx from this API is safe to treat as terminal — retrying it will produce the same answer — and any transient service-side failure arrives as 5xx with type \"api_error\" instead","audit":"every response carries a request_id, and a 5xx is logged server-side against that id with the underlying cause, so an error a customer saw can be attributed to a service fault or a genuine business answer after the fact","regression":"TestAPIKeyAuthDBFailureIsNot401, TestRequireTenantSplitsFaultFromProvisioning, TestLookupFailureIsNot404, TestAbsentRecordIsStill404, TestDBFailureIsNotErrNotFound"}},{"date":"2026-09-08","category":"fixed","summary":"Screening no longer reports a shortened lists_screened when a sanctions list's version could not be confirmed — the request now fails with 503 coverage_unavailable instead","details":"lists_screened and list_versions report which sanctions lists a screening actually ran against, and examiners read them as the proof of coverage. Previously, if one list's version could not be confirmed while the request was in flight, that list was quietly dropped from both fields and the response still returned 200 with a clean verdict — a narrower screen than was delivered, indistinguishable on the wire from a legitimately narrow one, with no error field anywhere to notice. The screening request now fails closed with 503 and error code coverage_unavailable rather than returning a partial answer as a complete one; the same applies to a point-in-time replay, and a batch job fails rather than persisting evidence rows that understate its coverage. A list that has no published snapshot at all is still omitted normally and is not an error — that list was genuinely not screenable, and lists_screened continues to widen on its own as lists come online. Retry a coverage_unavailable response; it is transient. No response shape changed, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"lists\":[\"ofac\",\"uk\",\"eu\"]}' -w '\\n%{http_code}\\n' | jq -c '{screened: .lists_screened, versions: [.list_versions[].list], code: .error.code}'","Repeat the call on a schedule and assert the invariant on every response: either the HTTP status is 200 and lists_screened contains every list you named, or the status is non-2xx. A 200 carrying only some of the lists you asked for is a failure."],"expect":"200 with lists_screened == [\"ofac\",\"uk\",\"eu\"] and list_versions carrying one entry per list; or, when a list version cannot be confirmed, 503 with error.code == \"coverage_unavailable\" and no matches or verdict in the body","audit":"every persisted screening carries the list_versions it ran against, so the stored record and the response agree on coverage by construction — a screening is never recorded against list versions it did not confirm","regression":"TestListVersionsFailsClosedOnLookupError, TestListVersionsSkipsMissingSnapshot, TestGetListVersionsAsOfPropagatesError"}},{"date":"2026-09-03","category":"changed","summary":"The CSV export now carries the jurisdiction advisory and the declared region, and the MCP screening tool surfaces an embargo nexus in its narrated result","details":"The CSV export projected the country a caller submitted but not the advisory it produced — the input rather than the finding, and the finding is the part that carries the obligation. Three columns are added: location_region (the sub-national region declared on the request), jurisdiction_risk (the jurisdiction code the advisory named), and jurisdiction_authority (its regulatory citation). They are blank for screenings with no nexus. The columns are appended within the existing row layout, so a consumer reading columns by NAME is unaffected; a consumer reading by fixed position must re-read the header row. The full JSON export is unchanged. Separately, the MCP screening tool now accepts region and city arguments so an agent can declare a sub-national nexus at all, keeps accepting a spelled-out country (\"Cuba\", \"IRAN\") alongside them rather than requiring an ISO code once a city is mentioned, and reports an embargo nexus in its summary line rather than only in the structured payload — an advisory on an otherwise clean name is exactly the finding a name-only reading of the result would miss. Batch results also gained a client-side shape, so a name screened in a batch and the same name screened singly no longer disagree about an embargo nexus depending on which surface was used.","endpoint":"/v1/export","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/export/bulk?format=csv\u0026from=\u003c7 days ago, YYYY-MM-DD\u003e\u0026to=\u003ctoday, YYYY-MM-DD\u003e\u0026limit=1\" -H \"X-API-Key: $KEY\" | head -1 | tr ',' '\\n' | grep -n 'location_region\\|jurisdiction'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"UA\",\"region\":\"UA-40\"}}' | jq -r .trace_id","curl -s \"https://api.noblesight.io/v1/export?format=csv\u0026trace_id=\u003ctrace_id from step 2\u003e\" -H \"X-API-Key: $KEY\" | tail -1 | tr ',' '\\n' | sed -n '11,13p'"],"expect":"step 1 lists columns 11, 12 and 13 as location_region, jurisdiction_risk and jurisdiction_authority; step 3 prints those three fields for the screening from step 2 as UA-40, UA-40, and \"E.O. 13685; 31 CFR Part 589\". Note the header comes from /v1/export/bulk, which pages a date range — /v1/export returns a single screening and requires trace_id or request_id, so it cannot produce a bare header row.","audit":"the flat export now records the embargo determination alongside the input that produced it, so a CSV pull is sufficient evidence of what was found rather than only what was asked","regression":"TestExportCSVRowJurisdiction, TestLocationArgKeepsLiberalCountry, TestClientMirrorsAPIResponseTypes"}},{"date":"2026-09-03","category":"added","summary":"/status now reports an embargo_table component: the comprehensive-embargo jurisdiction set degrades visibly when it goes too long without being reconciled against published OFAC programs","details":"Every other component on /status watches a machine process — a feed stops arriving, a poll fails, the database stops answering. The comprehensive embargo set has no feed to watch: programs are established and revoked by executive order, so the set goes wrong when nobody reads the Federal Register, and that failure is silent by nature. /status now carries embargo_table, which reports degraded once the least recently reviewed entry passes the reconciliation window, and degrades the overall status with a message naming the cause. Screening is unaffected either way — a degraded embargo_table means the set may no longer match current law, not that screening has stopped. GET /v1/jurisdictions reports each entry's last review date so the specific overdue entry is identifiable. The check reads an in-process table, so it needs no database and keeps reporting when other components cannot.","endpoint":"/status","verification":{"steps":["curl -s https://api.noblesight.io/status | jq '{overall: .status, embargo: .components.embargo_table.status}'","curl -s https://api.noblesight.io/v1/jurisdictions -H \"X-API-Key: $KEY\" | jq '[.jurisdictions[].reviewed_at] | min'"],"expect":"step 1 returns an embargo_table status of \"operational\" or \"degraded\" (never absent); step 2 returns the oldest review date across the set, which is the date the component measures — if embargo_table is degraded, that date is older than the reconciliation window","audit":"the reconciliation date is published per jurisdiction, so a reviewer can establish when each determination was last checked against published programs rather than taking currency on trust","regression":"TestEmbargoTableComponent, TestEmbargoTableNotAlreadyOverdue, TestJurisdictionReviewAgeUsesOldest"}},{"date":"2026-09-03","category":"added","summary":"New GET /v1/jurisdictions returns the comprehensive-embargo record, queryable at a past date — so a jurisdiction that was embargoed and later revoked stays answerable","details":"Comprehensive embargo programs are established and revoked over time, and there was previously no way to ask what was embargoed when. GET /v1/jurisdictions returns every jurisdiction on record with its program, regulatory citation, the dates the embargo took effect and was revoked, and when the entry was last reconciled against OFAC's published programs; GET /v1/jurisdictions/{code} returns one, and accepts every token screening accepts — the canonical code in any case, plus region names such as CRIMEA, SEVASTOPOL and DPRK — so a token that raises an advisory always has a record to answer for it. Both accept as_of and report whether the embargo was in force at that instant, so a bank reviewing a historical transaction can establish the status that applied at the time. effective_from is the effective date of the instrument that ESTABLISHED the embargo — the executive order or proclamation, never the CFR part that later implemented it, which for Cuba are sixteen months apart — and determination names the instrument behind every date on the record, so no published date rests on nothing. Two properties are worth stating. The record reports the LAW, which is deliberately a different question from what a screening was told: a screening performed while the engine's table was behind a revocation keeps the advisory it was issued, and this endpoint reports the revocation — both answers are preserved rather than one being corrected to match the other. And jurisdictions that were evaluated and determined NOT to be comprehensively embargoed are published as explicit entries with the reason cited, because an examiner asking why a region is not flagged deserves a determination rather than silence. Requires an API key; available on every tier.","endpoint":"/v1/jurisdictions","verification":{"steps":["curl -s \"https://api.noblesight.io/v1/jurisdictions/SY?as_of=2024-01-01\" -H \"X-API-Key: $KEY\" | jq '{embargoed, effective_to}'","curl -s \"https://api.noblesight.io/v1/jurisdictions/CU?as_of=1962-06-01\" -H \"X-API-Key: $KEY\" | jq '{embargoed, effective_from, determination}'","curl -s \"https://api.noblesight.io/v1/jurisdictions/SY?as_of=2026-01-01\" -H \"X-API-Key: $KEY\" | jq '{embargoed, effective_to, determination}'","curl -s https://api.noblesight.io/v1/jurisdictions -H \"X-API-Key: $KEY\" | jq '[.jurisdictions[] | select(.embargoed)] | length'"],"expect":"step 1 returns embargoed true; step 2 returns embargoed true with effective_from \"1962-02-07\" and a determination citing Proclamation 3447; step 3 returns embargoed false with effective_to \"2025-07-01\" and a determination citing E.O. 14312; step 4 returns 7, the jurisdictions under a comprehensive embargo today","audit":"each entry carries the regulatory citation establishing its status and the date it was last reconciled against OFAC's published programs, so a historical status can be traced to the instrument that set it","regression":"TestJurisdictionsEndpointBothTimelines, TestJurisdictionsEndpointCollection, TestJurisdictionsEndpointAcceptsScreeningAliases, TestEveryPublishedDateCitesItsInstrument, TestCubaEmbargoRunsFromProclamation3447, TestEmbargoJurisdictionMatchesRegistry"}},{"date":"2026-09-03","category":"added","summary":"New optional location object on /v1/screen and /v1/batch declares a sub-national nexus, so a party in Crimea, Sevastopol, or the covered DNR and LNR regions now raises a jurisdiction advisory","details":"The comprehensive embargoes covering Crimea, Sevastopol and the so-called DNR and LNR regions apply to regions, not countries, and those regions have no ISO 3166-1 country code. Ukraine itself is not embargoed, so a party in Sevastopol could only be described as country \"UA\" — correct and insufficient — and no advisory could fire. Requests may now carry location: {country, region, city}, where region takes an ISO 3166-2 subdivision code (for example \"UA-40\") or a common region name; the same object is accepted per item on batch. Detection reads the region first, then the location's country, then the existing country field, stopping at the first that resolves to a jurisdiction on record — so {country: \"UA\", region: \"UA-40\"} reports Sevastopol rather than unembargoed Ukraine, and {country: \"IR\", region: \"IR-07\"} still reports the Iran embargo, because an ordinary province defers to its country instead of suppressing it. Supplying more precise geography never removes a signal. Two guarantees: location is evaluated for the advisory only and never affects match scores, so declaring a nexus can never penalize your own results; and location.country must be an ISO 3166-1 alpha-2 code, rejected with invalid_parameter naming the field rather than accepted and ignored. Region is deliberately not checked against the embargo set — an ordinary region such as \"California\" is recorded and simply returns no advisory, because failing a screening, or a ten-thousand-name batch job, over a supplementary signal would be the worse outcome. A subdivision code that contradicts the country declared alongside it is still refused. The existing country field is unchanged and keeps working exactly as before. A nexus that is not declared is still not detected — names are never parsed for region words. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"UA\",\"region\":\"UA-40\",\"city\":\"Sevastopol\"}}' | jq '{j: .jurisdiction_risk.jurisdiction, a: .jurisdiction_risk.authority, l: .jurisdiction_risk.level}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"UA\"}}' | jq '.jurisdiction_risk'","curl -s -o /dev/null -w '%{http_code}\\n' -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"location\":{\"country\":\"IRN\"}}'"],"expect":"step 1 returns jurisdiction \"UA-40\" citing E.O. 13685 at level \"comprehensive_embargo\"; step 2 returns null because Ukraine itself is not embargoed; step 3 returns 400 with code invalid_parameter and param \"location.country\", because \"IRN\" is not an ISO 3166-1 alpha-2 code","audit":"the declared location is persisted with the screening request in the immutable record, so the reviewer of a flagged screening can see the nexus that was disclosed rather than inferring it from the country alone","regression":"TestScreenLocationRegionNexus, TestScreenLocationRegionFallsThroughToCountry, TestScreenLocationRejectsContradiction, TestValidateLocation, TestNexusTokenPrecedence"}},{"date":"2026-09-03","category":"fixed","summary":"Replaying a past screening now reproduces the jurisdiction advisory exactly as it was issued, instead of recomputing it against today's embargo programs","details":"The jurisdiction advisory is derived from a table of comprehensive embargo programs that changes when programs are established or revoked. Replay previously recomputed the advisory from the current table, so a screening performed before a program change replayed with the current answer rather than the one it was given — the record appeared to disagree with itself, and a replay could report a mismatch for a screening that was in fact reproduced faithfully. Replay is now pinned to the instant of the original screening and returns the advisory that was actually issued then, including one issued under a program that has since been revoked. Re-reading a completed batch is pinned the same way, so an item renders what was recorded at the time it was screened. The distinction is between two different questions, and each surface now answers the one it is asked. Replay asks what you were told, so a screening from before the advisory existed replays with no advisory rather than a retroactive one, and a screening issued under a program since revoked replays with the advisory it was given. An as_of screening asks what was true of the world, so it reports the embargo that was legally in force on that date: screening an Iranian counterparty as of 2020 returns the 31 CFR Part 560 advisory, because Iran has been comprehensively embargoed since 1995 — the age of our own table is not part of that answer. Re-reading a completed batch answers the same question and is pinned the same way. Each table entry now records when its program took effect and when it was revoked, so a jurisdiction that was embargoed in the past is preserved as history rather than deleted, and a replay's field-by-field comparison now includes the advisory itself: a replay that disagreed with the record about an embargo previously still reported verdict \"match\". Live screening behaviour, request and response shapes are unchanged — no version change.","endpoint":"/v1/export/replay","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zzyxx Cleanname\",\"country\":\"IR\"}' | jq '{trace: .trace_id, j: .jurisdiction_risk.jurisdiction, a: .jurisdiction_risk.authority}'","curl -s -X POST https://api.noblesight.io/v1/export/replay -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"trace_id\":\"\u003ctrace_id from step 1\u003e\"}' | jq '{verdict: .verdict, j: .replayed.jurisdiction_risk.jurisdiction, a: .replayed.jurisdiction_risk.authority}'"],"expect":"step 1 returns jurisdiction \"IR\" with authority \"31 CFR Part 560\"; step 2 replays the same screening with verdict \"match\" and an identical jurisdiction and authority — the advisory is reproduced from the record, not recomputed","audit":"the advisory is written into the screening's immutable record at decision time, and replay resolves the embargo table as it stood at that instant, so a decision made under a program later revoked still reproduces as it was made","regression":"TestJurisdictionKnowledgeTimeline, TestJurisdictionLegalTimeline, TestJurisdictionRiskKnownAt, TestJurisdictionInForceIsNotFlooredAtShipDate, TestScreenAsOfEmbargoInForce, TestReplayEndpointPreservesLocationNexus, TestCompareReplayJurisdictionDrift, TestDetectJurisdictionZeroTimeMatchesLive"}},{"date":"2026-08-31","category":"fixed","summary":"Ownership determinations are no longer replaced by a smaller set when the underlying ownership data is incomplete — the previous determinations stay in service instead","details":"The ownership determination set is rebuilt from scratch on a schedule. If the underlying ownership data was incomplete at the moment of a rebuild, the rebuild would previously still be published, replacing a larger set of determinations with a smaller one — so a company that returned blocked could quietly begin returning unknown, with no error anywhere to notice. The rebuild now compares itself against the last published set and refuses to publish when the number of determinations collapses, leaving the previous determinations serving and recording the refusal. The conservative direction is deliberate: continuing to serve slightly stale determinations can only over-report, which surfaces as a reviewable alert, while publishing an incomplete set under-reports invisibly. A genuine large reduction can still be published deliberately, and that decision is recorded. No request or response shape changed.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Advanced Mining Works\"}' | jq '{status: .ownership.status, pct: .ownership.cumulative_percent, as_of: .ownership.as_of}'","Repeat the identical call after the next scheduled rebuild has run (at least 24h later) and compare the two responses field by field."],"expect":"ownership.status is \"blocked\" on both calls with the same cumulative_percent; the status never changes to \"unknown\" or \"clean\" unless a sanctions list itself changed","audit":"the determination carries as_of and source_publish_ids pinning the data snapshot it was computed from, so a change in answer is always traceable to a change in source data","regression":"TestRecomputeShrinkGuard, TestShrankTooFar"}},{"date":"2026-08-31","category":"fixed","summary":"Ownership determinations now cover companies identified only in a sanctions list's narrative text — screening such a company returns its determination and evidence chain instead of unknown","details":"Some companies are never designated in their own right and appear only inside another entry's narrative description, which is where an ownership relationship to a designated person is often stated. Those companies are part of the ownership picture, but a portion of them could not be reached by name at screening time, so screening them returned unknown even though the relationship had been recorded. They are now reachable, and screening returns the determination the ownership computation reached for them, with the evidence chain and the source quote behind it. Two changes produce this: the recorded ownership assertions can now be restated into the ownership graph directly, so a determination is never stranded by an interrupted batch run; and a batch run is no longer treated as covering a document until that document's results are durable, so an interruption is retried rather than silently skipped. Expect some companies that previously returned unknown to return review or blocked. Response shape is unchanged and all four status values remain as documented, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Nomenko\"}' | jq '{status: .ownership.status, chains: (.ownership.chains | length)}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Nomenko\"}' | jq '.ownership.chains[0].hops[] | select(.is_sanctioned) | .name'"],"expect":"ownership.status is \"review\" (not \"unknown\"), with at least one evidence chain; the second call names the designated owner the chain reaches","audit":"each chain hop carries the entity it passed through and whether that entity is itself designated, and the determination carries as_of and source_publish_ids pinning the computation snapshot","regression":"TestReplayRepairsUnreachableDeterminations, TestExtractionWatermarkRequiresDurability"}},{"date":"2026-08-31","category":"fixed","summary":"Ownership determinations now reach names they previously missed, and clean is no longer returned when the ownership computation did not actually evaluate the subject — those cases return unknown","details":"One real-world company can appear in the ownership graph under several names, and the determination attaches to only one of them. Screening a different form of the name could therefore return clean while a determination existed under a sibling name. Three corrections: (1) when candidate names tie, the one carrying a determination is now preferred, so the determination is returned instead of being masked; (2) when the subject resolves to an entity the computation does not evaluate — one that is itself designated, and so directly blocked rather than derivatively blocked — the response is unknown rather than clean; (3) when a near-identically named entity carries a determination that the resolved one does not, the result is ambiguous and returns unknown rather than clean. A determination is never transferred onto a name it was not computed for, and a lower-similarity name never outranks a better one. Expect some subjects that previously returned clean to return unknown, and some to return review or blocked with full evidence chains. Response shape is unchanged and all four status values remain as documented, so no version is minted.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Naftiran Intertrade Company SRL\"}' | jq '{status: .ownership.status, pct: .ownership.cumulative_percent}'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Advanced Mining Works\"}' | jq '{status: .ownership.status, chains: (.ownership.chains | length)}'"],"expect":"the first no longer reports ownership.status \"clean\" for a subject whose determination exists under a fuller form of the same name; the second returns \"blocked\" with at least one evidence chain and a cumulative_percent at or above 50","audit":"the determination carries as_of and source_publish_ids pinning the computation snapshot, and a blocked result raises an alert retrievable via GET /v1/alerts","regression":"TestOwnershipResolver"}},{"date":"2026-08-31","category":"changed","summary":"Ownership determination semantics clarified: clean means no sanctioned ownership was found within the ownership graph's source coverage — advisory input to an institution's 50 Percent Rule program, not an exhaustive all-clear","details":"The OwnershipMatch.Status contract now states this coverage boundary explicitly, and the public capability description (llms.txt) carries the same limitation language in place of an unqualified capability claim. unknown continues to mean the name is not in the graph at all. No request or response shape changed and no field behavior changed — this aligns the public claims with the contract so both tell one story. Additive documentation correction; no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s https://api.noblesight.io/llms.txt | grep -A1 'Coverage boundary'","curl -s -X POST https://api.noblesight.io/v1/screen -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"\u003cany name\u003e\"}' | jq .ownership.status"],"expect":"the llms.txt coverage-boundary bullet contains the limitation sentence (clean = no sanctioned ownership found within the graph's source coverage, advisory not a substitute), and ownership.status returns one of blocked/review/clean/unknown exactly as the contract defines them"}},{"date":"2026-08-29","category":"added","summary":"Team registry and analyst identity: POST/GET/DELETE /v1/team manages the analysts alert assignment validates against, and the new X-Actor header attests which registered analyst performed an alert action — making four-eyes closure of high-score alerts workable for teams sharing one API key","details":"Alert assignment has required the assignee to be a registered active team member, but no endpoint existed to register one — assignment rejected every assignee, and because four-eyes closure of high-score alerts (score 90 and above) requires a second analyst identity that could never be introduced, such alerts could not be closed by anyone once reviewed. The team registry closes both gaps. POST /v1/team registers an analyst {email, role: analyst|admin} (idempotent — re-registering reactivates), GET /v1/team lists the registry, DELETE /v1/team/{email} deactivates without deleting so historical case activity stays resolvable. Any alert-mutating call may now carry an X-Actor header naming a registered active analyst; the action's audit actor is then \"analyst:\u003cemail\u003e\" instead of the key principal \"apikey:\u003cprefix\u003e\", and an unverifiable X-Actor is rejected with 400 invalid_actor rather than recorded. Four-eyes compares actors, so two analysts on one key satisfy it by each passing their own X-Actor. Additive — no version change.","endpoint":"/v1/team","verification":{"steps":["curl -s -X POST \"$BASE/v1/team\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"email\":\"analyst2@bank.example\"}' | jq '{email, role, status}'","curl -s -X POST \"$BASE/v1/alerts/$ALERT_ID/escalate\" -H \"X-API-Key: $KEY\" -H \"X-Actor: analyst1@bank.example\" -d '{}' | jq '.status'","curl -s -X POST \"$BASE/v1/alerts/$ALERT_ID/resolve\" -H \"X-API-Key: $KEY\" -H \"X-Actor: analyst2@bank.example\" -d '{\"status\":\"closed_false_positive\",\"reason\":\"false_positive_name_only\"}' | jq '.status'"],"expect":"registration returns the active member; a high-score alert escalated under one analyst's X-Actor closes under a second analyst's X-Actor — the same analyst attempting closure still receives four_eyes_violation, and an unregistered X-Actor receives 400 invalid_actor","audit":"every alert action records its actor in the immutable case activity as analyst:\u003cemail\u003e (validated against the registry at request time) or apikey:\u003cprefix\u003e, so an examiner can attribute each status change to a specific verified analyst; registrations and deactivations are logged with trace IDs","regression":"TestFourEyesUnblockedByXActor walks the full deadlock scenario end to end — escalate as one analyst, closure blocked for the same analyst, closure succeeds for the second, both identities in the audit trail; TestTeamRegistryAPI locks registration, idempotent reactivation, listing, validation, and deactivation"}},{"date":"2026-08-29","category":"fixed","summary":"Batch screening alerts now reach your alert queue: the alert_id on a batch item is retrievable via GET /v1/alerts/{id} and the alert appears in GET /v1/alerts — previously batch alerts were created outside your tenant's scope and could never be retrieved","details":"Alerts raised by batch screening items were created without the tenant association that every alert surface filters by, so the alert_id a batch item reported referenced an alert that GET /v1/alerts/{id} answered with 404 and GET /v1/alerts never listed — the batch channel was silently disconnected from alert disposition. Batch jobs now carry their tenant from creation and every alert they raise inherits it, and previously orphaned batch alerts have been reattached to their owning tenant where the owning key still exists, so historical batch alerts appear on the alert surfaces retroactively. Batch alerts follow the same creation rule as screening alerts: created when the item's highest score is \u003e= your alert_threshold. Alert queues can grow for customers who ran batches before this fix; response shapes are unchanged — no version change.","endpoint":"/v1/batch","verification":{"steps":["curl -s -X POST \"$BASE/v1/batch\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"names\":[{\"full_name\":\"Vladimir Putin\"}]}' | jq '.id'","curl -s \"$BASE/v1/batch/status?id=$JOB_ID\" -H \"X-API-Key: $KEY\" | jq '.items[0].alert_id'","curl -s \"$BASE/v1/alerts/$ALERT_ID\" -H \"X-API-Key: $KEY\" | jq '{id: .id, source: .source, status: .status}'","curl -s \"$BASE/v1/alerts?source=batch\" -H \"X-API-Key: $KEY\" | jq '[.data[].id] | index('$ALERT_ID') != null'"],"expect":"the alert_id the batch item reports resolves via GET /v1/alerts/{id} to an alert with source \"batch\", and the same alert is present in the GET /v1/alerts queue — never a 404 on an ID the API itself just returned","audit":"each batch item's immutable screening evidence record is unchanged; the alert it raises is now retrievable and disposable through the standard alert lifecycle with the full case-activity audit trail, closing the gap where batch hits could never be dispositioned","regression":"TestBatchAlertReachesQueue runs the real batch pipeline against a synthetic listed entry and proves the item's alert_id is tenant-scoped, retrievable by ID through the API, and listed in the tenant's queue"}},{"date":"2026-08-29","category":"fixed","summary":"GET /v1/alerts now returns every alert as created: raising your alert_threshold no longer retroactively hides existing lower-score alerts — including open ones mid-investigation — from the list endpoint","details":"The tenant alert_threshold gates alert creation exactly once, at creation time. The list endpoint additionally re-filtered reads by the CURRENT threshold, so tightening it made previously created lower-score alerts — open ones included — vanish from GET /v1/alerts even though the records themselves are immutable and remained retrievable by ID. An examiner asking \"show me every alert you generated\" now gets creation-time reality regardless of later configuration changes. Callers who want a read-time floor pass the existing min_score filter explicitly. Listings can grow for tenants that raised their threshold after alerts existed; response shapes are unchanged — no version change.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"$BASE/v1/alerts?limit=100\" -H \"X-API-Key: $KEY\" | jq '[.data[].score] | min'","curl -s \"$BASE/v1/account\" -H \"X-API-Key: $KEY\" | jq '.features.alert_threshold'"],"expect":"alerts created before a threshold increase remain in the listing, so the minimum listed score can be below the current alert_threshold; nothing that GET /v1/alerts/{id} can retrieve is absent from the list","audit":"alert records are append-only and were never deleted by threshold changes; this fix makes the primary retrieval surface reflect that stored reality, so list output, per-ID retrieval, and the immutable audit trail all agree","regression":"TestListAlertsSurvivesThresholdRaise creates an alert under a low threshold, raises the threshold above its score, and proves the alert still appears in GET /v1/alerts"}},{"date":"2026-08-29","category":"fixed","summary":"Typed identifier screening: an id type is now matched by case-insensitive prefix against publisher category strings, industry tokens like \"IMO\" are accepted, and a type matching no category returns a structured 400 — a typed request can no longer return a silent zero-match","details":"Screening by identifier with an explicit type previously required the caller's type string to equal the category string the publisher records — but publishers suffix their categories (\"Tax ID No.\", \"National ID No.\") and label vessel IMO numbers \"Vessel Registration Identification\", so a request like ids:[{number:\"8511823\",type:\"IMO\"}] filtered every row and returned a clean zero-match response while the untyped form of the same number matched. Type matching is now case-insensitive prefix against the stored category strings, common industry tokens are translated to the publisher's label (\"IMO\" reaches vessel registration identifiers), and a type that matches no stored category on any screened list is rejected with a structured 400 (code invalid_parameter, param naming the offending field) instead of screening nothing. Batch items with an unmatchable type are marked errored rather than reported as cleanly screened. Typed and untyped forms of the same identifier now return identical matches, and identifier numbers and types are bounded at 200 characters (real identifiers sit well under 100; longer values are rejected with a structured 400). Recall rises for typed requests; response shapes are unchanged — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"ids\":[{\"number\":\"8511823\",\"type\":\"IMO\"}]}' | jq '{count: .match_count, top: .matches[0].name, score: .matches[0].score}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"ids\":[{\"number\":\"8511823\",\"type\":\"NO-SUCH-CATEGORY\"}]}' | jq '{code: .code, param: .param}'"],"expect":"the typed IMO request returns the same deterministic score-100 vessel match as the untyped form \"IMO 8511823\"; the unmatchable type returns HTTP 400 with code invalid_parameter and param ids[0].type — never a zero-match success","audit":"typed identifier screenings persist to the same immutable audit record as every screening, and rejected types produce no screening record at all — an all-clear on file always reflects identifiers that were actually searched","regression":"TestScreenByIDTypedMatchesUntyped locks typed/untyped equivalence across token, label, and prefix forms on synthetic fixtures; TestScreenTypedIDHTTP locks the structured 400 with the exact offending parameter; TestScreenVesselByIMO keeps the untyped published-form pathway green"}},{"date":"2026-08-29","category":"changed","summary":"AI-generated name variations now cover every OFAC SDN entry — transliterations, romanization-system variants, and cultural equivalents are matched on every screening, refreshed daily as the list changes","details":"Screening has always consulted stored AI-generated name variations as a match pathway alongside primary names and OFAC's own aliases, but production coverage was partial. Every live OFAC SDN entry now carries a generated variation set — common English-press spellings, standard romanization systems (McCune-Reischauer and Revised Romanization for Korean, Wade-Giles and pinyin for Chinese, ALA-LC and press renderings for Arabic and Persian), and cultural equivalents — each stored with a written reason, so a match via this pathway is examiner-explainable, never a black box. A daily job generates variations for newly designated entries as list updates land. Reverse screening (portfolio monitoring) probes the same variation set, so both matching directions benefit. Scores for affected transliterations rise; response shapes are unchanged — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Osama bin Laden\",\"lists\":[\"ofac\"]}' | jq '{alert: .alert_id, top_sources: [.matches[0].match_sources[] | {source, matched_text}]}'"],"expect":"the common English spelling of a listed name returns a strong match whose match_sources include an ofac_ai_variation pathway with the matched variation text — alongside the primary-name and alias pathways","audit":"every stored variation carries a reason string persisted with it, and each screening's immutable audit record includes the variation pathway and matched text exactly as returned, so an examiner can trace any variation-driven match to its stored justification","regression":"TestScreenFoldTransliteration and the corpus battery lock transliteration recall; the daily variations-backfill job is resumable and skips covered entries, so coverage is monotonic"}},{"date":"2026-08-28","category":"fixed","summary":"Transliteration recall: name variants that differ in their first letter — Kasem/Qasem, Chong/Jong, Osama/Usama — now receive full phonetic credit instead of scoring below their canonical spellings","details":"Classic Soundex encodes every letter of a name into a phonetic class except the first, which it compares literally. That blind spot penalized standard romanizations of the same name that open with a different letter — the Arabic qaf rendered as K or Q (Kasem/Qasem), Korean Ch/J romanization systems (Chong/Jong), and O/U vowel openings (Osama/Usama). Two scoring paths were affected: such variants missed the phonetic corroboration the engine grants equivalent-sounding names, and multi-part names could be mistaken for single-token partial matches and discounted accordingly. Both paths now compare names with the first letter mapped through the same phonetic classes Soundex applies to every other letter, so a screening for a common English rendering of a listed name scores in line with the canonical spelling. Scores for affected variants rise; response shapes are unchanged — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Kasem Solaimani\"}' | jq '{status: .status, top: .matches[0].score, alert: .alert_id}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Qasem Soleimani\"}' | jq '{status: .status, top: .matches[0].score, alert: .alert_id}'"],"expect":"the transliterated and canonical spellings of the same listed name both return a strong match against the same designation and both raise an alert — the first-letter variant no longer scores materially below the canonical form","audit":"every screening's persisted record carries the scores as returned at decision time, so pre-fix and post-fix decisions each reflect the scoring in force when they were made and remain reproducible via point-in-time replay","regression":"TestScreenFoldTransliteration proves a first-letter variant of a synthetic listed name clears the alert threshold without tripping the partial-name discount while a genuine surname-only query still takes it; TestSoundexFoldMatch and TestApplyNameTokenSignal lock the phonetic equivalence table both ways"}},{"date":"2026-08-16","category":"added","summary":"OFAC 50 Percent Rule screening: /v1/screen responses now carry an ownership determination — an entity majority-owned by sanctioned parties surfaces as blocked with its evidence chains and raises its own alert, even when the entity appears on no sanctions list","details":"Under OFAC's 50 Percent Rule (FAQs 398/401), an entity owned 50% or more in the aggregate by blocked persons is itself blocked even though it is not listed. Screening now answers that question: every live name screening includes an ownership object with a determination — blocked (aggregate sanctioned ownership of 50% or more, with the ownership chains as evidence), review (a relationship to a sanctioned party exists but no source discloses a percentage; requires investigation), clean (known ownership structure, no sanctioned ownership found), or unknown (the name is not in the ownership data; absence of data is reported explicitly, never silently). Determinations are precomputed from sanctions-list ownership assertions corroborated by GLEIF corporate-relationship data, and each chain hop names its source, so the evidence is examiner-traceable; as_of screenings and identifier-only requests omit the field. A blocked determination raises a distinct alert (ownership_alert_id, alert source \"ownership\") that bypasses the score threshold — the determination is categorical, not a match confidence. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"$NAME\"}' | jq '{status: .ownership.status, pct: .ownership.cumulative_percent, chains: .ownership.chains, alert: .ownership_alert_id}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"$NAME\",\"as_of\":\"2026-08-01\"}' | jq '.ownership'"],"expect":"a live screening returns an ownership object whose status is blocked, review, clean, or unknown — blocked determinations carry cumulative_percent \u003e= 50, the evidence chains with per-hop sources, and an ownership_alert_id; the as_of screening returns null ownership (historical inquiries never carry current ownership data)","audit":"the ownership determination is persisted in the screening's immutable audit record exactly as returned, and each determination pins the source snapshots it was computed from (as_of plus source_publish_ids), so an examiner can reproduce the evidence behind any blocked decision","regression":"TestScreenOwnershipEndToEnd proves a name on no list but majority-owned by sanctioned parties returns status blocked with chains, raises the ownership alert, and lands in the audit record; TestOwnershipResolver locks all four determinations; TestOwnershipRecompute locks the FAQ 401 chain semantics behind the precomputation"}},{"date":"2026-08-15","category":"changed","summary":"The jurisdiction_risk table is reconciled against current OFAC programs: Syria no longer returns a comprehensive-embargo signal following the July 2025 revocation of its sanctions program, and every remaining entry carries its current program citation","details":"Executive Order 14312 (June 30, 2025) revoked the Syria sanctions program effective July 1, 2025, and OFAC removed the Syrian Sanctions Regulations (31 CFR Part 542) from the Code of Federal Regulations on August 26, 2025 — so a country input of Syria no longer produces an advisory jurisdiction_risk signal. Remaining Syria-related U.S. authorities are list-based designations, which name screening continues to cover. The comprehensive-embargo set is now Cuba (31 CFR Part 515), Iran (31 CFR Part 560), North Korea (31 CFR Part 510), Crimea and Sevastopol (E.O. 13685; 31 CFR Part 589), and the so-called DNR and LNR covered regions (E.O. 14065; 31 CFR Part 587). Kherson and Zaporizhzhia remain excluded because the Secretary of the Treasury has never determined them to be E.O. 14065 covered regions (OFAC FAQ 1006). Response shapes are unchanged — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"$NAME\",\"country\":\"SY\"}' | jq '.jurisdiction_risk'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"$NAME\",\"country\":\"IR\"}' | jq '{jurisdiction: .jurisdiction_risk.jurisdiction, program: .jurisdiction_risk.program, authority: .jurisdiction_risk.authority}'"],"expect":"Syria returns null jurisdiction_risk (no comprehensive embargo to signal); Iran still returns the advisory signal with its program and citation (31 CFR Part 560) — as do Cuba, North Korea, and the embargoed Ukrainian regions","audit":"each screening's persisted record includes the jurisdiction_risk signal (or its absence) as returned at decision time, so pre- and post-reconciliation decisions each reflect the programs in force when they were made","regression":"TestDetectJurisdiction locks the reconciled table both ways — the three country programs and four Ukrainian regions fire with complete program citations, while Syria post-revocation and the never-determined Kherson/Zaporizhzhia regions stay silent"}},{"date":"2026-08-15","category":"changed","summary":"UK screening now sources the FCDO UK Sanctions List — the UK government's single source for all UK designations since the OFSI Consolidated List closed in January 2026 — broadening UK coverage to every designation type","details":"The United Kingdom consolidated its sanctions publications: the OFSI Consolidated List of Asset Freeze Targets closed on 28 January 2026, and the UK Sanctions List published by the Foreign, Commonwealth \u0026 Development Office is now the sole authoritative source. UK screening ingests that list directly. Coverage broadens from asset-freeze targets to all UK designations (including travel-ban and other measures) — roughly 6,300 designations versus roughly 5,100 previously. Designation identifiers are unchanged (the same UKSL reference IDs, e.g. RUS0001), so alerts, match sources (uk_designation, uk_alias), and screening history remain continuous. GET /v1/lists now reports the list as \"UK Sanctions List (FCDO)\" with the publisher's own generation date as publish_date. Response shapes are unchanged — no version change.","endpoint":"/v1/lists","verification":{"steps":["curl -s \"$BASE/v1/lists\" | jq '.lists[] | select(.list==\"uk\") | {display_name, publish_date, record_count, imported_at, status}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"$NAME\",\"lists\":[\"uk\"]}' | jq '{lists_screened, list_versions}'"],"expect":"the uk list reports display_name \"UK Sanctions List (FCDO)\", a 2026 publish_date, a record_count around 6,300, and a fresh imported_at; screening restricted to uk returns the pinned uk list version alongside any matches — designations carrying only non-asset-freeze measures (e.g. travel bans) now match","audit":"every screening decision persists the exact UK list version it ran against, and the cutover import is recorded in the list's changelog and archive tables, so pre- and post-migration decisions are both reproducible against their respective snapshots","regression":"TestUKFreshLoadDeltaRoundTrip imports the full FCDO list and proves a clean zero-change re-import; TestParseXML and TestParseXMLFailClosed lock the new format's parsing, date normalization, and fail-closed identity guards; TestParseXMLLiveFixture asserts structural invariants against the live feed"}},{"date":"2026-08-15","category":"added","summary":"Canada (SEMA) and Australia (DFAT) consolidated sanctions lists are live — both appear in GET /v1/lists and participate in screening and portfolio monitoring","details":"Screening coverage now includes Canada's Special Economic Measures Act consolidated list and Australia's DFAT consolidated list. Both lists ingest on the standard polling cadence, appear in GET /v1/lists with their publisher snapshot and freshness status, are selectable via the screening lists parameter (canada, australia), and feed portfolio monitoring like every other source. Additive — no version change.","endpoint":"/v1/lists","verification":{"steps":["curl -s \"$BASE/v1/lists\" | jq '[.lists[] | select(.list==\"canada\" or .list==\"australia\") | {list, status, record_count, imported_at}]'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"$NAME\",\"lists\":[\"canada\",\"australia\"]}' | jq '.lists_screened'"],"expect":"both lists appear with a record_count in the thousands and a recent imported_at; screening with the lists parameter restricted to canada/australia returns matches attributed to those sources","audit":"every screening decision persists the sanctions list versions used, so records show exactly which Canada/Australia publisher snapshot each result was screened against","regression":"TestCanadaFreshLoadDeltaRoundTrip imports the full live-format SEMA list and proves a clean re-import; TestParseXMLRejectsDuplicateSids and TestBuildScheduleItem lock the entry-identity model against the live feed's format; TestDownloadLive (network-gated) proves the DFAT fetch end to end"}},{"date":"2026-08-15","category":"added","summary":"GET /status now reports sanctions-list ingestion freshness as a lists component — monitoring dashboards see degraded the moment any list's ingestion goes stale, without polling /v1/lists per list","details":"The public status endpoint gains components.lists, computed from the same freshness verdict as GET /v1/lists: operational when every deployed list was polled within the documented freshness window (24h), degraded when any list is stale. A degraded lists component degrades the overall status (HTTP 200 with status \"degraded\") — screening remains available and continues against the last imported lists; GET /v1/lists identifies which list is affected. When freshness cannot be determined the component fails closed rather than reporting health. Additive — no version change.","endpoint":"/status","verification":{"steps":["curl -s \"$BASE/status\" | jq '{status, lists: .components.lists}'","curl -s \"$BASE/v1/lists\" | jq '[.lists[] | {list, status}]'"],"expect":"the status response includes components.lists; when every list in /v1/lists reports \"ok\" the component is \"operational\" and overall status \"operational\" — if any list reports \"stale\", the component and overall status read \"degraded\" while screening endpoints keep responding","audit":"both endpoints derive from the persisted per-list ingestion heartbeat, so the status shown to a dashboard at any moment is reproducible from the recorded list versions and their poll timestamps","regression":"TestStatusListsComponent drives a list's ingestion heartbeat stale and proves /status degrades and then recovers with the heartbeat; TestFreshnessContractMatchesListhealth locks /status, /v1/lists, and the internal staleness alarm to one shared freshness computation"}},{"date":"2026-08-14","category":"fixed","summary":"Batch screening now fails closed on alert creation: an item whose alert cannot be recorded finishes with status error instead of screened, so a match can never complete silently without its alert","details":"Previously, if creating the alert for an above-threshold batch match failed, the failure was only logged internally — the item still finished as screened, with the match visible in its results but no alert for analysts to adjudicate and no error status to trigger a retry. Now an alert-creation failure marks the item error; re-submit the name to retry (alert de-duplication absorbs any partial state). A screened item with an above-threshold score always carries its alert_id. Additive — no version change.","endpoint":"/v1/batch","verification":{"steps":["curl -s \"$BASE/v1/batch/status?id=$JOB_ID\" -H \"X-API-Key: $KEY\" | jq '[.items[] | select(.status==\"screened\" and .highest_score\u003e=80 and ((.alert_id // 0)==0))]'"],"expect":"an empty array — no completed batch contains an item that screened at or above your tenant's alert threshold (default 80) without a linked alert_id; items whose alert could not be recorded appear with status \"error\" and can be re-submitted","audit":"batch item statuses and their alert links are part of the persisted batch record; an error item leaves no orphaned screened result, so the batch report and the alert queue can never disagree about which matches await review","regression":"TestBatchAlertFailureFailsItem forces an alert-creation failure through the real batch path and proves the item lands in error status with no alert link, while a clean run screens and links its alert"}},{"date":"2026-08-14","category":"fixed","summary":"An unrecognized country code no longer reduces match scores — absence of country knowledge is never treated as a country mismatch — and roughly sixty additional country codes (including XK, TL, BT) are now recognized","details":"Screening accepts an optional ISO 3166-1 alpha-2 country to refine scoring. Previously, a code outside the recognized set — newer or user-assigned assignments such as XK (Kosovo), TL (Timor-Leste), or BT (Bhutan) — could be treated as a mismatch against the listed party's countries and reduce the match score, even when the subject's country genuinely matched the sanctions record. A reduced score can keep a true match below the alert threshold, so this failed in the dangerous direction. Now a country code the engine cannot map contributes nothing: no penalty, no country dismissal signal — identical to omitting the country. The recognized code set also grew by roughly sixty countries, which now participate in country comparison normally. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"$NAME\",\"country\":\"XX\"}' | jq '{score: .summary.top_match.score, country_signals: [.matches[].dismissal_signals[]? | select(.field==\"country\")]}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"$NAME\"}' | jq '.summary.top_match.score'"],"expect":"both requests return the same top-match score and the unrecognized code produces no country dismissal signals; a recognized code (for example US against a Russian national) still applies the documented country handling","audit":"the persisted screening record preserves the request country and every dismissal signal, so an examiner can confirm which country comparisons ran for any historical decision","regression":"TestCompareCountryUnmappedCode locks that unknown codes never penalize while recognized-code mismatches still do; TestIsoToCountry locks the expanded code set"}},{"date":"2026-08-14","category":"fixed","summary":"Portfolio monitoring now screens sanctions-list changes under their stored AI name variations — a monitored name matching a new designation only through a transliteration or spelling variant raises an alert, with recall matching point-in-time screening","details":"Previously, point-in-time screening (POST /v1/screen) matched names against AI-generated variations of listed names (transliterations, spelling variants, cultural equivalents), but portfolio monitoring re-screened list updates against enrolled names using only each changed entry's name and aliases — a monitored customer whose name resembles a new designation only through such a variant was caught if screened directly, but a later list update did not raise an alert. Now every OFAC SDN list change is also screened under its stored variations, scored identically to the equivalent direct screen, so both directions catch the same matches. A variation match is fully attributed: the alert's match payload names the matched variant and the reason it was generated, and the persisted monitoring evidence records each variation screened. Additive — no version change.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"$BASE/v1/alerts/$ALERT_ID\" -H \"X-API-Key: $KEY\" | jq '.match_payload[0].Sources[0] | {Variation, Reason, MatchedText}'"],"expect":"for a monitoring alert raised because a list change matched an enrolled name through a stored variant, Variation holds the variant text that matched and Reason states why it was generated (for example a transliteration); for primary-name and alias matches both fields are empty, and MatchedText always shows which listed text was screened","audit":"each variation screened by monitoring persists its own evidence in the same append-only screening record store as point-in-time screens — including the variant text and its generation reason — for the 10-year retention window","regression":"TestReverseScreenAIVariationParity proves a variation-only match alerts identically through point-in-time screening and portfolio monitoring, with equal scores and the variant attributed on the alert; TestReverseScreenVariationExpansionOnlySDN locks which list's changes expand"}},{"date":"2026-08-14","category":"fixed","summary":"Portfolio-monitoring alerts now record every subsequent sanctions-list event on the open alert's timeline, and a designation on a different list raises its own alert instead of merging into the open one","details":"Previously, while a monitoring alert was open, further list events for the same person — a re-designation or a modified entry with new identifiers — deduplicated into the existing alert without leaving any record, and a same-named designation arriving from a different sanctions list merged into the open alert the same way, so an analyst could resolve an alert without ever seeing that a second list had also designated the person. Now every absorbed list event is appended to the alert's immutable activity timeline as a list_event entry (actor system:monitoring, with change type, list, publish id, and score in the note), visible in GET /v1/alerts/{id}; and the monitoring dedup key includes the source list, so a designation on a different list creates a distinct alert requiring its own adjudication. Duplicate open alerts for identical repeats of the same (entity, person, list) are still prevented. Additive — no version change.","endpoint":"/v1/alerts","verification":{"steps":["curl -s \"$BASE/v1/alerts/$ALERT_ID\" -H \"X-API-Key: $KEY\" | jq '[.activity[] | select(.action==\"list_event\")]'"],"expect":"for a monitoring alert whose subject appears in a later list update, the activity timeline contains a list_event entry with actor \"system:monitoring\" and a note naming the change type, source list, and publish id; if the same person is designated on a second sanctions list while the alert is open, GET /v1/alerts shows a separate open alert for that list rather than one merged alert","audit":"list_event entries live in the same append-only case-activity trail as status changes and notes — they cannot be updated or deleted, and they export with the alert's full timeline","regression":"TestMonitoringDedupRecordsListEvent proves an absorbed list event appears on the open alert's timeline via the API; TestCrossListDesignationCreatesDistinctAlert proves a second list's designation raises its own alert while same-list repeats still deduplicate"}},{"date":"2026-08-13","category":"fixed","summary":"deep_screen responses now report expansion coverage: a new deep_screen_coverage object counts variations generated vs screened, and any shortfall sets deep_screen_fallback_reason — partial enhanced recall is never silent","details":"Previously deep_screen_fallback_reason was set only when the AI variation expansion failed entirely; if individual generated variations could not be screened, the response reported success while delivering less recall than requested. Now every deep_screen response where variation generation succeeded carries deep_screen_coverage with variations_generated and variations_screened, and whenever screened \u003c generated the response also sets deep_screen_fallback_reason to variation_screening_failure. Equal counts positively confirm the full expansion was delivered; a whole-expansion failure still reports llm_failure. Retry a partial-coverage request to re-attempt full recall. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"deep_screen\":true}' | jq '{deep_screen_coverage, deep_screen_fallback_reason}'"],"expect":"deep_screen_coverage is present with variations_generated equal to variations_screened on a fully delivered expansion (deep_screen_fallback_reason absent); if any generated variation could not be screened, variations_screened is lower and deep_screen_fallback_reason is \"variation_screening_failure\"; if the expansion could not run at all, deep_screen_fallback_reason is \"llm_failure\" and no coverage object appears","audit":"the persisted screening record (GET /v1/export?trace_id=) preserves the full response, including the coverage counts and any fallback reason, for the 10-year retention window","regression":"TestDeepScreenCoverage proves every generated-vs-screened combination yields visible counts and an explicit fallback on any shortfall; TestScreenDecodesDeepScreenCoverage locks the client-side decode"}},{"date":"2026-08-13","category":"fixed","summary":"Webhook delivery now covers portfolio-monitoring alerts and retries durably: every alert source pushes, retry state survives restarts, and retries span roughly 10 hours instead of 31 seconds","details":"Previously webhooks fired only for alerts created by POST /v1/screen and POST /v1/batch — an alert raised by portfolio monitoring (a sanctions-list update matching an enrolled entity) was visible only by polling GET /v1/alerts. Delivery also retried for only ~31 seconds in process memory, so an endpoint down for a routine deploy permanently missed the delivery, and a server restart could drop pending batch deliveries. Now every created alert — api, batch, and monitoring — is enqueued to a durable outbox at creation time and delivered with up to 8 passes on a persisted backoff spanning roughly 10 hours. Retry state survives restarts; an endpoint that already accepted an event is never re-POSTed while other endpoints retry; and every attempt carries the same evt_* identifier in both the payload id and the X-Noble-Event-ID header (these previously disagreed), so consumers can deduplicate at-least-once delivery. A delivery failure never suppresses the alert — GET /v1/alerts remains the system of record. Additive — no version change.","endpoint":"/v1/webhooks","verification":{"steps":["curl -s -X POST \"$BASE/v1/webhooks\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"url\":\"https://your-endpoint.example.com/hooks\",\"secret\":\"a-shared-secret-at-least-32-chars-long\"}' | jq '{id, status}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq '{alert_id}'"],"expect":"within seconds your endpoint receives an alert.created event for the returned alert_id whose payload id equals its X-Noble-Event-ID header; if your endpoint is down it receives the same event (same id) on a later pass — including after a Noble server restart — rather than losing it; alerts created by portfolio monitoring arrive at the same endpoint with source \"monitoring\"","audit":"every attempt is recorded per endpoint with status, HTTP response code, attempt count, and error; the alert remains in GET /v1/alerts regardless of delivery outcome","regression":"tests/webhook_delivery_test.go — TestReverseScreenMonitoringDeliversWebhook drives a sanctions-list delta through the reverse-screen job to a signed delivery; TestWebhookOutboxRetriesAcrossRestart proves retry state survives a process restart"}},{"date":"2026-08-13","category":"fixed","summary":"Webhook notifications are enabled on the hosted deployment: POST /v1/webhooks registers endpoints and alert delivery is active (Standard tier and above)","details":"The hosted deployment previously rejected webhook registration with a 503 (\"webhooks not configured\") because the delivery subsystem was not configured. Registration and delivery now work as documented: HMAC-SHA256-signed payloads verified via X-Noble-Signature, endpoint secrets encrypted at rest, automatic retries with exponential backoff, and a per-attempt delivery record. A webhook failure never suppresses an alert — the alert queue remains the system of record. Additive — no version change.","endpoint":"/v1/webhooks","verification":{"steps":["curl -s -X POST \"$BASE/v1/webhooks\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"url\":\"https://example.com/hooks/noble\",\"secret\":\"a-shared-secret-at-least-32-chars-long\"}' | jq '{id, url, status}'","curl -s \"$BASE/v1/webhooks/stats\" -H \"X-API-Key: $KEY\" | jq '{total, remaining, limit}'"],"expect":"registration returns a webhook id with status \"active\" (not a 503 \"webhooks not configured\" error) for a Standard-tier-or-above key, with the URL masked in the response; stats shows the registration counted against the per-tenant limit","audit":"every delivery attempt for the registered endpoint is recorded with status, HTTP response code, attempt count, and error — webhook delivery records are part of the tenant's audit surface"}},{"date":"2026-08-13","category":"fixed","summary":"Coverage documentation corrected to match GET /v1/lists: OFAC Consolidated (Non-SDN) is live, headline counts now say 8 live lists, and Canada and Australia are shown as planned","details":"The docs page, marketing page, and llms.txt had drifted from the API in both directions: OFAC Consolidated (Non-SDN) is imported and screened but was still labeled planned, while headline counts claimed 10 live lists even though Canada (SEMA) and Australia (DFAT) are planned and do not yet appear in GET /v1/lists. All surfaces now state what the API states: 8 lists live — OFAC SDN, OFAC Consolidated (the two share the single ofac entry in GET /v1/lists), UK, EU, UN, France, Belgium, Netherlands — with Canada and Australia planned. Documentation only — screening behavior, scoring, and API responses are unchanged. Additive — no version change.","endpoint":"/v1/lists","verification":{"steps":["curl -s \"$BASE/v1/lists\" | jq -r '.lists[].list'","curl -s \"$BASE/v1/lists\" | jq -r '.lists[] | select(.list==\"ofac\") | .display_name'"],"expect":"exactly seven entries — belgium, eu, france, netherlands, ofac, uk, un — with neither canada nor australia present, and the ofac entry's display_name is \"OFAC SDN + Consolidated\"; this matches the coverage table on the docs page (8 live rows, with OFAC SDN and OFAC Consolidated listed separately, and Canada and Australia marked Planned)"}},{"date":"2026-08-06","category":"added","summary":"Point-in-time replay: POST /v1/screen accepts as_of to screen against the sanctions lists as they stood at a past instant, and POST /v1/export/replay re-runs a stored screening pinned to its original list state and reports a field-level comparison verdict","details":"Previously the audit trail recorded what the engine decided, but nothing could prove the decision reproduces: the temporal schema existed with no code path to screen against a historical snapshot. Now: (a) POST /v1/screen accepts as_of (RFC 3339, or YYYY-MM-DD meaning end of day UTC) — the screening evaluates against list rows effective at that instant, the response carries screened_as_of and the as-of list_versions, and the request is persisted to the audit trail but never creates alerts, enrolls the entity, or dispatches webhooks; (b) POST /v1/export/replay (Standard tier and above, client-isolated) loads a stored screening by trace_id or request_id, re-runs the exact compute path pinned to the record's screened_at, and returns verdict match|mismatch with field-level differences and explicit caveats for what cannot deterministically replay (tenant watchlist matches, deep_screen LLM variation influence, OFAC child-table detail). Scope: api-channel records only — batch and monitoring records return replay_unsupported_channel. Additive — no version change.","endpoint":"/v1/export/replay","verification":{"steps":["TRACE=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq -r '.trace_id')","curl -s -X POST \"$BASE/v1/export/replay\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d \"{\\\"trace_id\\\":\\\"$TRACE\\\"}\" | jq '{verdict, differences, caveats}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"as_of\":\"2026-08-01\"}' | jq '{screened_as_of, list_versions}'"],"expect":"the replay returns verdict \"match\" with no differences for a just-recorded screening (its list state has not changed); the as_of request returns screened_as_of \"2026-08-01T23:59:59.999999999Z\" (a date-only as_of means end of day UTC) and list_versions resolved to the imports active at that instant, not today's","audit":"GET /v1/export?trace_id=$TRACE shows the stored record replay compared against; an as_of screening's audit record persists the historical request verbatim, so the replay chain is itself replayable","regression":"tests/replay_test.go — inserts a list entry, screens, mutates the list, and asserts replay reproduces the original decision with the original list version"}},{"date":"2026-08-06","category":"added","summary":"GET /v1/alerts/resolutions/stats: observed false-positive rate by score band, aggregated from analysts' structured resolution reason codes","details":"Every alert resolution carries a structured reason code; this endpoint aggregates those dispositions into a measured FP rate per score band — the production counterpart to the fixture-derived FP benchmark, and the driving input for tuning your alert_threshold. Counting model: resolution events from the immutable case-activity trail joined to the alert's score; a reopened-and-reclosed alert contributes each closure, because the rate measures analyst decisions, not current alert states. Optional from/to (YYYY-MM-DD, inclusive) bound the resolution date. Additive — no version change.","endpoint":"/v1/alerts/resolutions/stats","verification":{"steps":["curl -s \"$BASE/v1/alerts/resolutions/stats\" -H \"X-API-Key: $KEY\" | jq '{total, false_positives, true_matches, observed_fp_rate, by_score_band}'"],"expect":"a disposition breakdown where observed_fp_rate equals false_positives/total and each by_score_band entry carries its own resolved/false_positives/true_matches/fp_rate; a tenant with no closed alerts gets zeroed totals, not an error","audit":"each counted disposition is an immutable case-activity row — GET /v1/alerts/activity?action=status_change returns the underlying events the stats aggregate","regression":"api/v1 resolution-stats handler tests + tests/ integration coverage landed in the same PR (#106)"}},{"date":"2026-08-06","category":"fixed","summary":"deep_screen/deep_summary no longer pin a retired Gemini model, and the audit trail records the exact serving model instead of the request alias","details":"The hardcoded gemini-2.5-flash pin 404s for newly created Gemini API keys, which surfaced as deep-screen fallbacks or 503 deep_screen_unavailable on affected deployments. AI calls now track the current stable Gemini model instead of a hardcoded pin, and summary.generated_by/model_version now report the exact serving model returned by the API rather than the requested alias — so the 10-year audit record names the model that actually produced the output. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"deep_summary\":true}' | jq '{generated_by: .summary.generated_by, model_version: .summary.model_version, fallback: .summary.fallback_reason}'"],"expect":"on an AI-configured deployment, generated_by is \"llm:\u003cserving-model-id\u003e\" with model_version naming a live model (never the retired gemini-2.5-flash); otherwise generated_by is \"deterministic\" with fallback_reason set — a 404-dead model can no longer be silently pinned","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e — the persisted summary block carries the same exact serving-model id","regression":"internal/gemini model-selection tests (#104)"}},{"date":"2026-08-04","category":"added","summary":"jurisdiction_risk: screenings whose country input touches a comprehensive OFAC embargo (Cuba, Iran, North Korea, Syria, Crimea/DNR/LNR regions) now return an advisory nexus signal on /v1/screen and per-item on /v1/batch","details":"Previously the country field fed only score penalties — it could lower a match score on mismatch but never raise a flag, so a request disclosing an Iran or Cuba nexus with a clean name returned no_significant_match and no signal at all. Now the same request also returns jurisdiction_risk: the canonical jurisdiction code (ISO 3166-1 alpha-2, or ISO 3166-2 for the embargoed Ukrainian regions UA-43/UA-40/UA-14/UA-09), jurisdiction name, OFAC program, regulatory citation, level (comprehensive_embargo), and fixed EDD guidance. Detection accepts alpha-2 codes, the Ukrainian region codes, and common English names; Ukraine (UA) and Russia (RU) themselves never fire — only the embargoed regions do. The signal is advisory nexus detection, not list matching: it is independent of matches, never alters scores or the summary verdict, and never blocks. It is persisted in the screening's immutable audit record and in each batch item's evidence record. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"John Cleanname\",\"country\":\"IR\"}' | jq '.jurisdiction_risk'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"John Cleanname\",\"country\":\"US\"}' | jq '.jurisdiction_risk'"],"expect":"the IR request returns a jurisdiction_risk block with jurisdiction \"IR\", program \"Iranian Transactions and Sanctions Regulations\", authority \"31 CFR Part 560\", level \"comprehensive_embargo\", and EDD guidance — even though the name itself has no significant match; the US request returns null. GET /v1/export?trace_id= for the IR screening shows the same block persisted in the audit record","audit":"GET /v1/export?trace_id=\u003ctrace_id from the IR screening\u003e — response_payload.jurisdiction_risk carries the identical block, proving the disclosed nexus is part of the 10-year record, not just the wire response"}},{"date":"2026-07-30","category":"fixed","summary":"deep_screen no longer degrades silently: an unconfigured deployment returns 503 deep_screen_unavailable, and a runtime AI failure sets deep_screen_fallback_reason on the response","details":"Previously a deep_screen request could silently fall through to standard screening with zero signal — no latency difference, no deep_screen match sources, no error — when the deployment had no AI backend configured or the AI call failed at runtime (#62). A paid enhanced-recall feature must never quietly deliver standard recall. Now: (a) a deployment without an AI backend rejects deep_screen requests with 503 deep_screen_unavailable (retry without deep_screen, or contact support) — deterministic misconfiguration fails hard; (b) a transient runtime AI failure still returns standard-recall results but sets deep_screen_fallback_reason on the response (values mirror the summary fallback_reason constants), so callers can detect and retry — the same honesty contract deep_summary has always had. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"deep_screen\":true}' | jq '{sources: [.matches[].match_sources[].source] | unique, fallback: .deep_screen_fallback_reason}'"],"expect":"exactly one of three honest outcomes: match sources include \"deep_screen\"/\"ofac_ai_variation\" (expansion ran), OR deep_screen_fallback_reason is set (transient AI failure, standard recall delivered and disclosed), OR the call itself fails 503 deep_screen_unavailable (deployment has no AI backend). A silent fall-through — no deep sources, no fallback reason, HTTP 200 — is the defect and can no longer occur"}},{"date":"2026-07-29","category":"added","summary":"The audit trail now covers every screening channel: batch items and portfolio-monitoring evaluations persist full evidence records, and /v1/export/bulk gains name= and channel= filters for entity-centric retrieval","details":"Previously only POST /v1/screen decisions were written to the immutable audit store — batch items and monitoring evaluations were invisible to GET /v1/export, and no entity-centric retrieval existed (an examiner's \"show me all screening activity for this customer\" could not be answered). Now: every batch item persists a screening_result evidence record including the list versions it screened against (batch previously captured none); every portfolio-monitoring evaluation persists a per-client evidence record in the same transaction as its alerts, and monitoring alerts link to their evidence via screening_result_id; export records carry two new fields — channel (api|batch|monitoring) and screened_name — and GET /v1/export/bulk accepts ?name= (case-insensitive, cross-channel entity filter) and ?channel=. The request/response payloads inside export records are the verbatim audit payloads; their shape is channel-dependent and documented on ExportRecord (api-channel records are byte-identical to before). POST /v1/screen is now fail-closed on audit persistence: a screening that cannot be recorded returns an error instead of an unrecorded success. Additive — no version change.","endpoint":"/v1/export/bulk","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' \u003e/dev/null","curl -s -X POST \"$BASE/v1/batch\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"names\":[{\"full_name\":\"Vladimir Putin\"}]}' \u003e/dev/null; sleep 3","curl -s \"$BASE/v1/export/bulk?from=$(date -v-1d +%F 2\u003e/dev/null || date -d yesterday +%F)\u0026to=$(date +%F)\u0026name=Vladimir%20Putin\" -H \"X-API-Key: $KEY\" | jq '[.data[].channel] | unique'"],"expect":"the export returns records for the same name from more than one channel (at least [\"api\",\"batch\"]); each record carries channel, screened_name, and its verbatim request/response payloads incl. list_versions"}},{"date":"2026-07-29","category":"fixed","summary":"DOB/country mismatch penalties now apply to matches from all 10 list sources — France, Belgium, Netherlands, UN, Canada, and Australia matches were previously exempt","details":"Secondary-attribute scoring (OFAC FAQ 5 Step 4: penalize a name match whose date_of_birth or country contradicts the list record) only fetched attribute data for OFAC, UK, and EU matches. Matches from the other six lists silently skipped the check, so a qualified request kept the full name-match score even when the supplied DOB or country disproved the match — inflated false positives on those lists. All prefixed list sources now route to source-specific attribute fetchers (with per-list date-format parsing, incl. Dutch dd-mm-yyyy and UN year-only records). Scores for mismatched-attribute matches on the six lists will DROP after this date; matches without stored attribute data remain unpenalized (absence of data never proves a mismatch). Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["# Pick any match from a non-OFAC/UK/EU list, then screen it with a contradicting DOB:","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"\u003cname from a france/un/canada match\u003e\",\"lists\":[\"france\"],\"date_of_birth\":\"1901-01-01\"}' | jq '.matches[0] | {score, dismissal_signals}'"],"expect":"when the list record carries a DOB, the score is reduced and dismissal_signals contains a date_of_birth entry (previously: full score, empty signals). If the record has no stored DOB, score is unchanged — absence of data is not treated as a mismatch"}},{"date":"2026-07-28","category":"fixed","summary":"Portfolio monitoring matches at the same sensitivity as /v1/screen, deduplicates alerts for API-key-only accounts, and processes list updates in minutes regardless of portfolio size","details":"Three monitoring fixes ship together. (1) Sensitivity parity: reverse screening (list update → your portfolio) previously matched at a stricter sensitivity than /v1/screen — a name pair the screening endpoint flags could be silently missed by monitoring. Both directions now match at identical sensitivity. Tenants with a low configured alert_threshold may see additional monitoring alerts after this date; those represent matches that were always intended to alert. (2) Dedup: monitoring alerts for accounts without a tenant workspace were never deduplicated — the same portfolio entity + sanctioned person could alert repeatedly across list imports. Deduplication now applies to all accounts, backed by a database uniqueness constraint. (3) Scale: reverse screening moved from the list-watcher hot path to a resumable background job with a trigram-indexed match probe — list updates are now screened against portfolios of any size in minutes, and an interrupted run resumes instead of restarting. Additive — no version change.","endpoint":"/v1/alerts","verification":{"steps":["# After the next sanctions list update that matches your portfolio:","curl -s \"$BASE/v1/alerts?status=new\u0026limit=100\" -H \"X-API-Key: $KEY\" | jq '[.data[] | select(.source==\"monitoring\")] | group_by(.portfolio_entity_id, .sdn_first_name, .sdn_last_name) | map(length) | max'"],"expect":"1 (or null if no monitoring alerts yet) — no open duplicate alerts exist for the same portfolio entity + sanctioned person; re-imports of the same list data do not re-alert"}},{"date":"2026-07-09","category":"fixed","summary":"/v1/batch now applies the same DOB/country/entity_type and watchlist scoring as /v1/screen — a name scores identically on both endpoints","details":"Completes the screen/batch scoring-parity work from #60 for qualified requests (#71). The #57/#60 fix unified the single-token partial-name penalty across both paths, but two other score-mutating steps still ran only on /v1/screen: the secondary-attribute penalties (screen.AdjustScores — DOB/country/entity_type, OFAC FAQ 5 Steps 2 \u0026 4) and the tenant watchlist. A batch item carrying a DOB or country qualifier therefore scored higher than the same name+qualifier via /v1/screen — the higher-stakes real-time path was the weaker one. Both endpoints now run one shared scoring pipeline (screen.FinalizeScores: watchlist → secondary-attribute penalties → single-token penalty), so a name scores identically whichever endpoint is called. /v1/batch names[] additionally accept entity_type, matching /v1/screen; an omitted entity_type never penalizes. Additive — no version change.","endpoint":"/v1/batch","verification":{"steps":["S=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"country\":\"US\"}' | jq '.matches[0].score')","ID=$(curl -s -X POST \"$BASE/v1/batch\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"names\":[{\"full_name\":\"Vladimir Putin\",\"country\":\"US\"}]}' | jq -r '.id')","sleep 2; B=$(curl -s \"$BASE/v1/batch/status?id=$ID\" -H \"X-API-Key: $KEY\" | jq '.items[0].highest_score'); echo \"screen=$S batch=$B\""],"expect":"the /v1/screen top score equals the /v1/batch highest_score for the qualified request (both reflect the country_mismatch penalty — screen == batch, not batch \u003e screen)","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e replays the /v1/screen result with its list_versions","regression":"tests/screen_batch_parity_test.go::TestScreenBatchQualifiedParity"}},{"date":"2026-07-06","category":"added","summary":"Weak-vs-strong AKA quality is now surfaced: match_sources[].alias_quality and matches[].weak_alias_only","details":"OFAC flags certain aliases as weak because they are a dominant false-positive source, and does not expect a weak-AKA-only hit to be treated like a strong-name match (FAQ 122-124). Previously a hit driven solely by a weak alias was indistinguishable from a strong-name hit. Each match_source that fired via an OFAC AKA now carries alias_quality (\"weak\" or \"strong\"; empty for primary-name, variation, ID, and non-OFAC pathways), and each match carries weak_alias_only — true when every pathway corroborating the match is a weak alias, so callers can deprioritize or auto-dismiss those hits. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"THE CHESS PLAYER\"}' | jq '.matches[] | select(.uid==\"4107\") | {weak_alias_only, sources:[.match_sources[]|{source,alias_quality}]}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq '.matches[0].weak_alias_only'"],"expect":"the weak-AKA hit (uid 4107) has weak_alias_only=true with an ofac_aka source of alias_quality \"weak\"; the Putin top match (primary name / strong alias) has no weak_alias_only","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e replays the result with its list_versions","regression":"tests/api_test.go::TestScreenAliasQuality; internal/nobleclient/methods_test.go::TestScreenDecodesRichMatchFields"}},{"date":"2026-07-06","category":"added","summary":"matches[].prohibition_type distinguishes a full SDN block from a CMIC investment ban, SSI sectoral restriction, or reject","details":"OFAC hits do not all carry the same action, but they were framed identically. An SDN (or 50%-Rule) listing is a blocking designation — freeze the property and file a blocking report — but a CMIC listing (E.O. 13959) is a securities investment ban, an SSI listing (E.O. 13662) is a sectoral dealing restriction, and FSE/NS-PLC are rejects; none of those are blocked. Each OFAC match now carries prohibition_type — one of blocking, cmic_investment, ssi_sectoral, fse_reject, plc_reject, capta — derived from its programs, so a caller applies the correct action instead of over-blocking a CMIC or SSI entity. Blocking dominates a mixed listing (an entity on both SDN and a sectoral program is blocking). Non-OFAC matches omit the field, since the taxonomy is OFAC-specific. Additive — no version change. (The SSI Directive number is a planned follow-up; it is not in the current ingest.)","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\"}' | jq '[.matches[] | {source, prohibition_type}]'"],"expect":"every OFAC match carries prohibition_type (an SDN listing is \"blocking\"); matches from non-OFAC lists omit prohibition_type","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e replays the result with its list_versions","regression":"api/v1/prohibition_test.go::TestProhibitionType; tests/api_test.go::TestScreenProhibitionType"}},{"date":"2026-07-06","category":"fixed","summary":"Screen a sanctioned vessel by its IMO number; identifier matching is now punctuation-insensitive on both sides","details":"Identifier screening normalized only the input (stripping spaces, dashes, dots, slashes) but compared it against the raw stored value, so any identifier published with punctuation could never match. The most visible casualty was vessel IMO numbers, which OFAC publishes as \"IMO 9102198\": a vessel was unscreenable by its permanent identifier. Stored identifiers are now normalized identically to the input (backed by new functional indexes across every list's ID table, so matching stays an index lookup), and a vessel matches whether screened as \"IMO 9102198\" or the bare \"9102198\". This also fixes the broader class of punctuated identifiers (hyphenated tax IDs, spaced registration numbers) across all lists. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"ids\":[{\"number\":\"IMO 9102198\"}]}' | jq '{verdict:.summary.verdict, top:.matches[0].last_name, type:.matches[0].sdn_type, score:.matches[0].score}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"ids\":[{\"number\":\"9102198\"}]}' | jq '.matches[0].last_name'"],"expect":"both the \"IMO 9102198\" and bare \"9102198\" forms return the vessel CLIPPER (sdn_type \"Vessel\") at score 100 with an alert, no name supplied","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e replays each screen with its list_versions","regression":"tests/api_test.go::TestScreenVesselByIMO"}},{"date":"2026-07-06","category":"added","summary":"Screen a digital-currency address or any identifier on its own — name is no longer mandatory when an identifier is supplied","details":"A wallet address, passport, tax ID, or other identifier can now be screened with no name: send it in ids (e.g. {\"ids\":[{\"number\":\"0x…\"}]}) and the listed entity is returned by exact ID match. A type is optional — when omitted the number is matched against every identifier category. An address pasted into the name field is now also caught: a whitespace-free alphanumeric token is additionally exact-matched as an identifier, so a listed wallet address returns its SDN instead of only sub-threshold fuzzy noise (matching OFAC's exact-match discipline for identifiers, FAQ 594). A request with neither a name nor an identifier is rejected with 400 query_required. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"ids\":[{\"number\":\"0x7FF9cFad3877F21d41Da833E2F775dB0569eE3D9\"}]}' | jq '{verdict:.summary.verdict, top:.matches[0].last_name, score:.matches[0].score}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"0x7FF9cFad3877F21d41Da833E2F775dB0569eE3D9\"}' | jq '.matches[0].score'","curl -s -o /dev/null -w '%{http_code}\\n' -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"limit\":5}'"],"expect":"the address (via ids, and pasted into name) returns GARANTEX EUROPE OU at score 100 with an alert; a request with neither name nor identifier returns HTTP 400 code query_required","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e replays each screen with its list_versions","regression":"tests/api_test.go::TestScreenIdentifierOnly"}},{"date":"2026-07-04","category":"fixed","summary":"Screening rank is now independent of limit; whitespace-only names are rejected","details":"Two fixes (#58). (1) The limit parameter caps how many matches are returned but must not change which match ranks #1. The internal top-entities selection truncated equal-scored rows in arbitrary order, so limit=1 could surface a different top match than limit=10 for the same query — a caller using limit=1 as \"the single best match\" could clear or dismiss against the wrong record. A uid tiebreak now makes the ranking a total order, so limit=N returns exactly the top N of one stable ordering (score descending, then uid) and the #1 match is identical for every limit. (2) A whitespace-only name (e.g. \"   \") was accepted and screened; it is now trimmed before validation and rejected with 400 name_required like an empty name. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["A=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Mohammed Ali\",\"limit\":1}' | jq -r '.matches[0].uid')","B=$(curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Mohammed Ali\",\"limit\":10}' | jq -r '.matches[0].uid'); [ \"$A\" = \"$B\" ] \u0026\u0026 echo RANK_STABLE","curl -s -o /dev/null -w '%{http_code}\\n' -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"   \"}'"],"expect":"the limit=1 and limit=10 top-match uid are equal (RANK_STABLE); the whitespace-only screen returns HTTP 400 (code query_required once identifier-only screening shipped, name_required before)","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e replays each screen with its list_versions","regression":"tests/screen_test.go::TestScreenLimitRankingStable; tests/api_test.go::TestScreenWhitespaceNameRejected"}},{"date":"2026-07-04","category":"fixed","summary":"entity_type is now a safe disambiguation hint — supplying the correct type never suppresses an alert, and the single-token penalty no longer misfires on organizations","details":"Two entity-recall bugs (#59). (1) The entity_type_mismatch penalty compared the caller's canonical entity_type against each list's raw, incompatible subject-type vocabulary (OFAC \"Entity\", France \"Personne morale\", EU \"enterprise\", UN \"entity\"), so supplying the documented entity_type:\"entity\" for an organization such as SBERBANK triggered a spurious penalty large enough to drop a confirmed match below the alert threshold. Both sides are now normalized to a canonical vocabulary before comparison; only a genuine, known-category mismatch is penalized, and an unknown or unmapped source type is never penalized — guaranteeing that supplying the correct entity_type can never lower a score relative to omitting it. (2) The single-token partial-name penalty (OFAC FAQ 5 Step 3, a person-name control) was applied to organizations, which are routinely referenced by one distinctive token (\"Sberbank\"); it now skips organizations, vessels, and aircraft. Additive scoring correction — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Sberbank\"}' | jq '{score:.matches[0].score, alert:.alert_id}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Sberbank\",\"entity_type\":\"entity\"}' | jq '{score:.matches[0].score, alert:.alert_id}'"],"expect":"both calls return matches[0].score \u003e= 80 with an alert; the entity_type=entity score is \u003e= the no-hint score (never lower)","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e replays each result with the list_versions that produced it","regression":"internal/screen/attributes_test.go::TestAdjustScoresEntityTypeNormalization + ::TestCanonicalEntityType; internal/screen/dismissal_signals_test.go::TestApplyNameTokenSignalSkipsOrganizations; tests/api_test.go::TestScreenEntityTypeRecall"}},{"date":"2026-07-04","category":"fixed","summary":"lists_screened now reflects the lists actually loaded and queried, not a static supported-list array","details":"The lists_screened field on every /v1/screen response was a static array of all supported list identifiers, so it could report coverage that is not ingested (e.g. Canada, Australia) and contradict the same response's list_versions. It is now derived from list_versions — the set of lists with a publisher snapshot actually on file — so lists_screened is always a subset of list_versions, and a list appears only once it is genuinely loaded and queryable. Additive — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Zxqwerty Nonexistent Person\"}' | jq '{screened:.lists_screened, versions:[.list_versions[].list]}'"],"expect":"lists_screened has exactly the same members as .list_versions[].list; it never contains a list absent from list_versions","audit":"GET /v1/sanctions-lists returns the authoritative loaded set — it must match lists_screened","regression":"tests/api_test.go::TestScreenListsScreenedMatchesLoaded"}},{"date":"2026-07-04","category":"fixed","summary":"Corrected the published /llms.txt evaluation script — step 3 no longer promises a verdict the API does not return","details":"The \"Evaluate Noble in 60 seconds\" script in /llms.txt promised that screening \"Vladimir Petrov\" returns summary.verdict: no_significant_match as proof of false-positive discipline; in reality that common name can reach a match and fire an alert — the opposite of the documented result. Step 3 now demonstrates discipline with a check that is true by construction: re-screen a known strong match with a contradicting attribute ({\"name\":\"Vladimir Putin\",\"country\":\"US\"}) and observe the score drop with a structured country_mismatch dismissal_signal that explains the reduction. Documentation-only correction — no behavior change.","endpoint":"/llms.txt","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Vladimir Putin\",\"country\":\"US\"}' | jq '{score:.matches[0].score, signals:.matches[0].dismissal_signals}'"],"expect":"matches[0].score \u003c 100 and dismissal_signals contains {reason_code:\"country_mismatch\", impact_on_score:-30}; the /llms.txt step-3 example now matches the observed behavior","audit":"GET /llms.txt step 3 describes exactly this call and outcome","regression":"internal/screen/dismissal_signals_test.go::TestCompareCountryStructured"}},{"date":"2026-07-04","category":"fixed","summary":"Name-variant recall: hyphen/space/transliteration variants no longer drop a confirmed match below the alert threshold; /v1/screen and /v1/batch now score identically","details":"The OFAC FAQ 5 Step 3 single-token partial-name penalty was misfiring on trivial orthographic variants because it compared name tokens by exact string equality. 'Kim Jong-un' (hyphen), 'Bashar Al Assad' (spacing), 'Qassem Soleimani' (doubled consonant), and 'Ramzan Kadirov' (i/y transliteration) each counted as a single-token match and had their score halved from 100 to 50 — below the default 80 alert threshold, producing no alert on a confirmed SDN (a false negative). The token check now splits on hyphens and uses the same Soundex equivalence as the core trigram+phonetic engine, so a variant the engine already scores as a full match is no longer penalized. Separately, the penalty ran only on the real-time /v1/screen path and not on /v1/batch, so the two endpoints returned different scores for the same name; both now run the identical scoring pipeline. Genuine single-token collisions (a bare surname such as 'Putin' against 'Vladimir Putin') are still penalized. Additive scoring correction — no version change.","endpoint":"/v1/screen","verification":{"steps":["curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Kim Jong-un\"}' | jq '{verdict:.summary.verdict, score:.matches[0].score, alert:.alert_id}'","curl -s -X POST \"$BASE/v1/screen\" -H \"X-API-Key: $KEY\" -H 'Content-Type: application/json' -d '{\"name\":\"Qassem Soleimani\"}' | jq '{score:.matches[0].score, alert:.alert_id}'","# parity: the single-name batch highest_score must equal the /v1/screen score for the same name"],"expect":"matches[0].score \u003e= 80, summary.verdict == \"confirmed_match\", and alert_id != null for each variant; and screen(name).matches[0].score == batch([name]).items[0].highest_score","audit":"GET /v1/export?trace_id=\u003ctrace_id\u003e replays the same result with the list_versions that produced it","regression":"internal/screen/dismissal_signals_test.go::TestApplyNameTokenSignal; tests/screen_batch_parity_test.go::TestScreenBatchVariantParity"}},{"date":"2026-06-17","category":"added","summary":"Confidential Portfolio (BYOK) — encrypt portfolio PII at rest under a customer-controlled key","details":"New Enterprise add-on. When enabled, a client's portfolio PII (name, DOB, country, external ref) and the matching monitoring-alert names are stored encrypted at rest under a per-customer KMS-wrapped key; the server decrypts in-memory only to run portfolio monitoring and to serve the owner's authenticated reads. Manage the key at /v1/portfolio/key: GET returns status (active/revoked/none), POST provisions, DELETE?confirm=\u003cclient_id\u003e revokes and crypto-shreds (irreversible — destroys the KMS key so the data is permanently unreadable). GET /v1/account now reports features.confidential_portfolio. This is encryption at rest with a customer-controlled key, not zero-knowledge: monitoring keeps running. Substring name search over encrypted entities is unavailable (exact dedup still works via a blind index). Additive — no version change.","endpoint":"/v1/portfolio/key"},{"date":"2026-06-07","category":"added","summary":"Stats companions for portfolio, webhooks, and case activity — uniform collection counts","details":"Counts are now a first-class read on every collection, not just alerts. New GET /v1/portfolio/stats (total + by_score_band, honors q/min_score/max_score), GET /v1/webhooks/stats (total, by_status, limit, remaining), and GET /v1/alerts/activity/stats (total + by_action, honors the same alert_id/actor/action/from/to filters as the list) join the existing GET /v1/alerts/stats. The convention is uniform: a collection at /v1/\u003cname\u003e exposes its counts at /v1/\u003cname\u003e/stats, and a stats endpoint honors the same filters as its list sibling. This recovers the row counts dropped from list envelopes in the 2026-06-01 cursor-pagination change, without re-adding O(n) total to page fetches. Additive — no version change.","endpoint":"/v1"},{"date":"2026-06-06","category":"added","summary":"Noble-Version header for API version pinning","details":"Clients may pin a dated API version with the Noble-Version request header (e.g., Noble-Version: 2026-06-01); the resolved version is echoed in the Noble-Version response header. No header pins to the latest version, so existing integrations are unaffected; an unrecognized version is rejected with 400 invalid_version. This entry is additive and therefore does not itself mint a new version — the latest version remains 2026-06-01 (cursor pagination), the most recent breaking change.","endpoint":"/v1"},{"date":"2026-06-01","category":"changed","summary":"Uniform cursor pagination across all list endpoints — {data, has_more}, ?after=\u0026limit=","details":"Every collection endpoint (alerts, portfolio, watchlist, export/bulk, alerts/activity, webhooks) now returns the same envelope: {\"data\": [...], \"has_more\": bool}. The array key is always \"data\". Pagination is keyset cursor only — ?after=\u003cid\u003e\u0026limit=\u003cn\u003e, ordered by id ascending; pass the id of the last item in data as ?after= for the next page. This is a breaking change to the wire shape: the per-endpoint keys (alerts, entities, entries, records, webhooks), offset/page params, and total counts are removed. Use GET /v1/alerts/stats for bounded counts; use the async export job for full extraction. /v1/webhooks returns the same envelope but is not paginated (capped per tenant).","endpoint":"/v1/alerts","breaking":true},{"date":"2026-05-27","category":"fixed","summary":"summary.fallback_reason guaranteed on every non-LLM deep_summary response","details":"Hardened the deep_summary path with a deferred invariant: when deep_summary=true and the response carries deterministic prose (LLM unavailable, marshalling error, Gemini failure, or empty LLM payload), fallback_reason is now guaranteed to be populated with llm_unavailable or llm_failure. Closes a gap where callers requesting a paid feature could receive deterministic prose with no signal that the LLM did not deliver. Empty LLM text or model from the Gemini path is now treated as llm_failure rather than passing through.","endpoint":"/v1/screen"},{"date":"2026-05-26","category":"added","summary":"match_sources[].matched_text and matched_length surface which string fired each pathway","details":"Every match-source row now carries the actual source-list string that produced its trigram score — the primary name, alias, or variation text the comparison hit. matched_length is the rune count, pre-computed for server-side aggregation. Lets consumers distinguish 'matched primary name Vladimir Putin' from 'matched 4-character alias JOHN', and is the foundation for length- and quality-aware scoring follow-ups. Empty when not applicable (e.g., exact ID matches).","endpoint":"/v1/screen"},{"date":"2026-05-26","category":"fixed","summary":"Filter OFAC metadata (Gender, Secondary sanctions risk:) out of top_match.ids and matches[].ids","details":"OFAC publishes some non-identifier features (gender, sanctions-program annotations) using the same \u003cID\u003e element as real identifiers. These were leaking into the ids array on every match and misleading compliance officers and AI agents. Only real identifier types (Passport, Tax ID, SWIFT/BIC, crypto wallet addresses, etc.) now surface; the rest are filtered at the API boundary.","endpoint":"/v1/screen"},{"date":"2026-05-26","category":"added","summary":"summary.fallback_reason surfaces when deep_summary silently falls back to deterministic","details":"When deep_summary=true is requested but the LLM path doesn't execute (server not configured, Gemini call failed/timeout), the response now carries fallback_reason (llm_unavailable | llm_failure) so the caller can tell that a paid feature did not deliver. Absent when deep_summary was not requested or when the LLM produced the prose successfully.","endpoint":"/v1/screen"},{"date":"2026-05-26","category":"added","summary":"Structured dismissal_signals on every match (reason codes, severity, score impact)","details":"Each penalty applied during attribute comparison now emits a structured signal alongside the legacy free-text reason. Stable reason codes enable filtering, aggregation, and audit-defensible dismissal records. The legacy dismissal_reasons field is preserved byte-for-byte during a 90-day deprecation window; both fields are guaranteed to have equal length and ordered correspondence.","endpoint":"/v1/screen"},{"date":"2026-05-26","category":"deprecated","summary":"dismissal_reasons (free-text array) — use dismissal_signals instead","details":"Free-text reasons cannot be filtered or aggregated. The structured dismissal_signals field carries the same information as reason codes, severity, and per-signal score impact. Removal scheduled with the next major API version; both fields remain populated until then.","endpoint":"/v1/screen"},{"date":"2026-05-26","category":"changed","summary":"Enforce OFAC FAQ 5 Step 3: single-token name match against multi-token sanctioned name is penalized","details":"Matches where only a single token of the request name overlaps with a multi-token sanctioned name (e.g., 'Putin' alone against 'Vladimir Putin') now receive a substantial score reduction and a partial_name_match_single_token dismissal signal. This codifies OFAC's 'just one of two or more names matching, i.e. just the last name' not-a-valid-match condition.","endpoint":"/v1/screen"},{"date":"2026-05-25","category":"added","summary":"GET /v1/changelog returns machine-readable API release notes","endpoint":"/v1/changelog"},{"date":"2026-05-25","category":"changed","summary":"POST /v1/screen accepts multiple ids; matches include matched_via","details":"Send an ids array (each with id_type + value). When a match is found by ID, the match object carries a matched_via field naming which ID hit.","endpoint":"/v1/screen"},{"date":"2026-05-25","category":"added","summary":"summary.top_match carries OFAC 5-step PII (DOB, addresses, IDs)","endpoint":"/v1/screen"},{"date":"2026-05-24","category":"added","summary":"Opt-in executive summary on screening responses (summary field)","endpoint":"/v1/screen"},{"date":"2026-05-15","category":"added","summary":"Idempotency-Key header supported on all state-mutating POST endpoints","details":"Retries with the same key within 24h return the cached response and produce zero side effects."},{"date":"2026-05-13","category":"fixed","summary":"meta.disclaimer and structured errors enforced across all v1 endpoints"},{"date":"2026-05-12","category":"added","summary":"Invite-only signup available via /v1/admin/invites"}],"entry_count":205,"fingerprint":"6b7eb53f254d6be150b18a4860b3c7e8a36c1be88df9dd6c8b1e7894f19c7469","meta":{"disclaimer":"Noble Sight provides sanctions screening results for informational purposes only. This service is not a substitute for a comprehensive sanctions compliance program. Noble Sight does not provide legal, regulatory, or compliance advice. Screening results reflect data available at the time of the request and may not capture all sanctions designations, aliases, or name variations. Final screening decisions, risk assessments, and compliance obligations remain the sole responsibility of the subscribing institution. Use of this service does not satisfy or replace any obligation under OFAC regulations, the Bank Secrecy Act, or any other applicable law."}}
