Sanctions Screening API - OFAC, EU, UN Consolidated Lists

Give it a list of customer or supplier names. Get back, for every name, whether it appears on the US Treasury OFAC SDN list, the EU financial sanctions list or the UN Security Council consolidated list — with the alias it matched, how close the match was, and which version of each list it was checked against.

Straight from the three government files. No third-party aggregator, no licence to worry about, no API key.

Why "per name", not "per hit"

Every other screening Actor in the store charges for the records it returns. Screen 10,000 suppliers, find 3 matches, and they bill you for 3 records — which sounds generous until you realise the work you actually paid for was the 9,997 names that came back clean, and that is the answer you needed for your audit file.

This one charges per name screened. A clean list costs exactly the same as a dirty one, because it took exactly the same work, and a clean row is a real answer. Rows that could not be screened — an empty input, or a run where no list could be downloaded — are not charged.

A clean name is an answer, not an empty cell

Every input name produces exactly one row:

Field	Meaning
query	The name you gave
hitCount	How many matches were found, 0 included. This is the number found, not the number returned
matches	The best maxMatches of them. If more were found, reason says so
topScore	Best match score, 0 when nothing matched
each match	listedName, matchedName, matchType, score, source, programs, countries, listedOn, reference
listVersions	The version of each list this answer is based on — e.g. {"ofac":"08/26/2026","eu":"2026-08-05T16:47:04+02:00","un":"2026-08-26T23:00:04Z"}
listsUsed / listsUnavailable	Which lists this answer actually covers, and which it does not

That last pair matters more than anything else here. A screening tool that quietly answers "no match" after failing to fetch a list is worse than no tool at all, because it produces a clean audit trail for a check that never happened.

• No list could be read → every row comes back ok: false, screened: false, hitCount: null, the failures are named, and nothing is charged.
• Some lists could be read → the row is answered, but reason names the lists it was NOT checked against. It never says "every list".

matchType says how it matched, not just that it did

Value	Meaning
exact-name / exact-alias	After normalisation, the strings are identical
reordered-name / reordered-alias	Same words, different order — HUSSEIN SADDAM against the alias SADDAM HUSSEIN
contained-name / contained-alias	Every word you typed appears in the listed name, which has one or two more — Nicolas Maduro against Nicolas MADURO MOROS
fuzzy-name / fuzzy-alias	Close but not identical. score says how close

A reordered match is never labelled exact, because that column ends up in an audit file.

What it catches

Measured on the real lists on 2026-08-27 — 19,319 OFAC + 6,234 EU + 1,011 UN records, 78,733 names including aliases:

You type	It finds	How
Saddam Hussein Al-Tikriti	21 records, the three exact-name ones first	exact name, then aliases and relatives
Abu Ali	27 records — it is a very common alias, and every one of them is returned with its own score	alias
HUSSEIN SADDAM	The same person	reordered
aero caribbean	AEROCARIBBEAN AIRLINES	alias, spacing
Sadam Husein	SADDAM HUSSEIN	misspelled, score 0.857
Bank Meli Iran	BANK MELLI IRAN, and the EU's Bank Melli	misspelled, both lists
김동운	KIM Tong-un	Korean script
Χαμάς	Hamas	Greek script
صدام حسين التكريتي	SADDAM HUSSEIN AL-TIKRITI	Arabic original script, from the UN list
Dorde Vukovic	matches records written Đorđe Vuković	Latin-extended letters folded
Nicolas Maduro	Nicolas MADURO MOROS	middle name missing
Toyota Motor Corporation	nothing	correctly clean
John Smith	nothing	correctly clean

Matching uses Levenshtein distance for spelling and a token overlap for word order, blocked on both whole words and three-letter word prefixes — which is what makes Sadam Husein reachable at all, since neither of its words appears in any list.

minScore is yours to set, between 0.5 and 1.0. 1.0 returns only identical strings and word-order variants of them. 0.85 is the default and catches the misspellings above. Below 0.8, false alarms start to outnumber real hits — at 0.5, most ordinary names match something. A value outside 0.5–1.0 falls back to the default rather than to the loosest setting, so a typo in your config cannot silently flood you with false positives.

What it does not do

Being honest about this is part of the product:

• A one-word query is not enough. Measured against 600 real listed entities and individuals on 2026-08-27, at the default threshold:

  What you type	Found
  The listed name as published	100%
  The listed name with one letter wrong	100%
  Surname first / word order swapped	100%
  A company name without its LTD / LLC / OAO suffix	93%
  A person without their middle name	84%
  Only the first two words of a long listed name	74%
  A single word (Hezbollah, Gazprom)	not matched at all

  A single word is deliberately never treated as "contained in" a longer listed name, because Ali alone would match dozens of unrelated people. Type the name as your records hold it — two words or more.

• Common personal names will hit. Screening 50 ordinary names on 2026-08-27 (30 large companies, 20 everyday personal names) produced 7 matches, every one of them a personal name — Wei Zhang, Ali Hassan, Muhammad Ali and similar. None of the 30 company names produced a false hit. A person-name hit is a candidate to review, not a finding.

• It screens names. It does not check dates of birth, passport numbers or addresses, so a common name will produce matches you must review by hand.

• It is not a full KYC or PEP check. There is no politically-exposed-persons list here, and no UK, Swiss, Canadian or Australian list yet. OFAC's SSI (non-SDN) list is not included either.

• It reads the lists fresh on every run, which takes about 40 seconds for all three (57 MB). It does not tell you what changed since your last run.

• A match is a candidate, not a legal determination. The reference on every match is the government's own record id, so you can open the source and decide.

Input

Field	Default	What it does
names	3 sample names	One name per line, people or companies. Rows that are not text are returned with a note, and not charged
sources	ofac, eu, un	Which lists to check. Repeats are ignored, and an id this Actor does not know is reported back to you rather than dropped
minScore	0.85	Match threshold, 0.5–1.0
maxMatches	10	Matches returned per name, best first. hitCount still reports the true total
maxRetries	3	Retries when a government server times out. A 404 is never retried

Sources

List	Publisher	Updated
SDN	US Department of the Treasury, OFAC	Almost daily — the copy read on 2026-08-27 was published 2026-08-26
Financial Sanctions List (FSF)	European Commission	On regulation changes
Consolidated List	UN Security Council	On committee decisions

All three are public government files, fetched directly. The exact version of each is written into every row, so an answer you filed six months ago can still be explained.

Other tools by NeverEmpty

Every NeverEmpty Actor follows the same rule: it never returns an empty result to mean two different things, and it only charges for rows that actually carry an answer.

• eu-tender-monitor - EU public tender notices from the official TED API
• domain-lookup - WHOIS, DNS, MX, SPF, DMARC and SSL expiry for a list of domains
• website-contacts - emails, phones and social profiles from any website, no invented addresses