OFAC Sanctions Screening: SDN and Consolidated Lists

Screen a name, or a batch of names, against the live United States Treasury OFAC sanctions lists. Every row states the verdict, any matched designations, and the exact list edition it was screened against.

This is a screening tool. A hit is a starting point for human review, not a compliance decision, and it is not legal advice. OFAC's own guidance is that apparent matches need verification.

Measured, not asserted

Against 240 real designations drawn from the live list, screened as OFAC spells them: 240 caught, 0 missed. Against 200 real registered US businesses: 0 flagged, everything came back clear.

Names are rarely typed the way a list spells them, so the harder question is what survives dirty data. Across 519 deliberately corrupted spellings of real designations, 93.4% were still caught:

how the name was wrong	caught
a character doubled	100.0%
a space dropped or added	100.0%
name parts reordered	96.6%
a character dropped	96.2%
a middle name or patronymic missing	93.3%
two characters transposed	90.0%
a vowel changed	78.8%

A changed vowel is the weakest case, because it alters the letters a comparison leans on without shortening the name enough to look like a typo. Short names are the other soft spot: once a name is only a few characters, one wrong character leaves too little signal to match on, and this tool would rather return clear than guess.

Every number above is reproducible. tests/measure_accuracy.py runs the whole measurement against the live lists with a fixed seed, and it fails loudly if its own controls do not hold, so a run that quietly screened nothing cannot report a perfect score.

What you get

One row per name screened:

query	verdict	topScore	matchCount	lei	listsScreened	retrievedAt
AEROCARIBBEAN AIRLINES	strong	1.0	1		CONSOLIDATED,SDN	2026-08-01T09:12:44Z
Aero-Caribbean	strong	1.0	1		CONSOLIDATED,SDN	2026-08-01T09:12:44Z
Acme Widgets Incorporated	clear	0.0	0		CONSOLIDATED,SDN	2026-08-01T09:12:44Z

Each match carries the OFAC uid, whether it hit a primary name or an alias, the subject type, the sanctions programs, any listed countries, and the list and publish date it came from.

A matched company also gets its Legal Entity Identifier when the official GLEIF register holds a record under exactly that name. Be aware this is uncommon: measured on 60 random entity designations it filled in 2, because OFAC publishes long official forms that no register lists verbatim. The check is deliberately strict. GLEIF's own name filter returns near matches, so two unrelated Mexican companies both came back as the same third company, and the lookup now discards any answer whose returned name is not the name asked for. An empty LEI field means no record was confirmed, never that none exists.

How it decides

Names are folded before comparison: accents removed, case and punctuation dropped, and for companies the corporate suffixes (LLC, GmbH, Pte, and the rest) set aside, so ACME TRADING LLC and Acme Trading compare as the same name. Aliases on the OFAC entry are matched as well as the primary name, which is how Aero-Caribbean finds a designation filed under AEROCARIBBEAN AIRLINES.

Three verdicts: strong, possible, and clear. possible exists on purpose. Near misses are surfaced for a human to look at rather than quietly cleared, because a missed designation costs far more than a second look.

What it costs

Billed per name screened. A clear result bills exactly like a hit, because knowing a counterparty is clear is the answer you came for. There is no actor start fee.

Two things are never billed. A duplicate name inside one batch is answered once and the repeat is free. And if the sanctions list fails to load, the run stops with an error and bills nothing, rather than returning a comforting all clear it cannot stand behind.

Names beyond maxNames are returned free in a separate dataset marked cap_reached, so nothing is silently dropped.

Your list is not kept

The names you submit are screened and returned. They are not written to any storage that outlives the run, and they are not retained, reused, or enriched into anything else.

Sources

• OFAC Specially Designated Nationals list, US Department of the Treasury
• OFAC Consolidated non-SDN lists (FSE, SSI, NS-MBS and others)
• GLEIF Legal Entity Identifier register, for company enrichment

All three are free public publications. Every row carries the source URL, the list publish date, and the time it was read.

Limits worth knowing

Matching is deterministic and name-based. It does not resolve identity from a date of birth, a passport number, or an address, so a common personal name can return more than one designation to review.

Cyrillic is transliterated so a Russian, Ukrainian or Belarusian record can reach the Latin list. Other scripts are not, and a name with nothing screenable left after normalisation is returned free in the rejected dataset marked unscreenable_name, never billed as a clear result.

Screening is only as current as the list edition stamped on the row. OFAC adds and removes names without a fixed timetable, so re-screen when it matters.