Counterparty Due Diligence in Seconds: What US Public Records Show, and What They Miss

We screened three companies across a dozen US government databases by API, in under ten seconds each. Here is what the records showed, and what they missed.

Why we ran this

Counterparty due diligence is the check you run before you sign with someone: are they who they say they are, are they sanctioned, are they barred from government work, who has a claim on their assets. Most guides on the topic explain what it is. Few show one being run.

So we ran one. We took a public API that puts US government records behind a single key and screened three companies chosen to produce three different results:

• IBM, a large listed company that should come back clean.
• Rosneft, the Russian state oil company, which should light up every sanctions list.
• Leprino Foods, a privately held Colorado company with no stock listing, where the interesting signals sit in less obvious records.

For each one we queried the same stack in the same order. Here is what that stack is, what it cost, and what surprised us.

The stack

A US counterparty check touches three kinds of record.

Who they are. The Global Legal Entity Identifier Foundation (GLEIF) index holds about 3.4 million Legal Entity Identifiers (LEIs), the 20-character codes that regulated firms use to identify counterparties. Each record gives the legal jurisdiction and the company's registration number there. State business registries (we used New York, Colorado and Connecticut, plus Florida's Sunbiz) confirm the entity exists and is in good standing. SEC EDGAR covers about a million filers.

Whether you may deal with them. The Treasury's OFAC lists name sanctioned people and companies. The Consolidated Screening List merges every US export and sanctions screening list across agencies into one search. SAM.gov lists parties excluded from federal contracts and grants.

What they owe. Uniform Commercial Code (UCC) filings are the public notices lenders file when they take a security interest in a company's assets. They tell you who gets paid first if things go wrong.

Every one of these, except Delaware's registry, answered from a maintained copy of the source in well under a second. Each call cost one credit on the provider's free plan, which grants 5,000 a month.

At that rate the free plan covers roughly 400 first-pass screens a month.

Five things the data taught us

1. SAM.gov double-counts OFAC

Our first read of the Rosneft results counted two independent red flags: OFAC and SAM.gov. That was wrong.

SAM.gov returned 13 exclusions for Rosneft. Every one of the 13 lists OFAC as the excluding agency. SAM.gov isn't recording a separate procurement ban here. It's republishing Treasury's designations inside the federal contracting system. A hit on both is one designation seen twice, not corroboration.

That matters if your screening process scores risk by counting hits. Count them naively and a single sanctions designation looks like a pattern.

2. One company, three ID systems

The six OFAC records for Rosneft aren't six companies. They're four. Two of them, the parent company and its trading arm, appear twice: once on Treasury's main sanctions list (the SDN list) and once on its Sectoral Sanctions Identifications list, under the same entity number and the same sanctions programs. The parent is entity 17022 on both.

The Commerce Department's Entity List, surfaced through the Consolidated Screening List, describes the same company under its own ID with a differently shaped record. Commerce's version is the tidier one: it has a clean start date (17 September 2014) and Federal Register citations in their own fields. OFAC's version packs the Russian tax number and directive references into one free-text remarks field.

Dedupe by name and you undercount distinct legal designations. Dedupe by ID and you count one company as several unrelated ones. Either way, the join logic is yours to write.

3. Most liens on file are dead

Leprino Foods has 26 UCC filings on record in Colorado. One is in force: filed on 11 October 2023, lapsing on 11 October 2028. The other 25 have lapsed or been terminated.

"26 liens" is a true count of filings and a badly misleading picture of the company's current secured debt. The field that matters is whether each filing is still in force, and it's easy to skip.

Retention also differs by state. Colorado keeps about 2.6 million filings going back to 1966, lapsed ones included. Connecticut keeps only filings in force or ended within the past year. An empty Connecticut result tells you far less than an empty Colorado one.

4. H-1B filings are a workforce signal hiding in plain sight

The Department of Labor publishes every H-1B labour condition application and permanent labour certification it has decided since fiscal year 2008, with worksites. Leprino Foods, with no stock listing and a bare-bones SEC record, has 282 H-1B applications on file. For comparison, Ping Identity, a Denver identity-software company, has 59.

For a private company that files little else, this is one of the few independent public reads on headcount, offices and seniority. It is also capped: IBM's count hit the search's 1,000-result ceiling, so for large filers you have to slice by year to get a real number.

5. State registries track existence, not ownership

Thoma Bravo agreed to buy Ping Identity and take it private. Ping's SEC filings stop on 28 October 2022. Its Colorado entities are still listed in good standing, with nothing in the state record that hints the company changed hands.

The record isn't wrong. A secretary of state records that an entity exists and has paid its fees. It doesn't track who owns it. If your question is "has control of this counterparty changed?", a clean state record answers a different question.

What this stack cannot see

Ten seconds is fast. It is not complete, and here is where it stops.

• Only US lists. This stack has no EU, UK or UN sanctions lists. A counterparty sanctioned in Brussels or London and not in Washington comes back clean. For anyone outside the US, that is the biggest gap.
• No beneficial ownership. None of these records tells you who ultimately owns or controls the company. The LEI index did turn up Rosneft subsidiaries in Germany, Luxembourg, Ireland, the UK and Switzerland, but only because they carry the Rosneft name. A subsidiary with a different name would not have surfaced.
• Name matching is literal. OFAC matches across aliases. Most registries don't. A counterparty trading under another name can slip past.
• Coverage is patchy. Four state registries plus Florida isn't fifty. Delaware, where most large US companies incorporate, isn't a maintained dataset in this API. It's a live lookup that costs ten times as much.
• Freshness varies. Most sources we touched were a day old. The SEC 13F holdings data was about two weeks behind.
• No adverse media, no politically exposed persons. Records tell you what the government has formally decided. They don't tell you what the press is reporting.

This is a first pass, and a good one. It won't do the whole job.

Our take

For decades the expensive part of a first-pass check was retrieval: knowing which database to open, working its search form, copying the result into a memo. That part has collapsed to seconds and cents.

What hasn't collapsed is judgment. Knowing that SAM.gov echoes OFAC. Knowing that 26 liens can mean one. Knowing that "good standing" says nothing about a buyout. The data got cheap. Reading it correctly didn't.

Fetching records faster is solved. Where we think legal AI earns its place is in reading them the way an experienced counsel would, and in saying plainly what the check did not cover. Courts are writing the same bar into case law: see how Latin American judges are drawing the line on AI.

Key Takeaways

• A first-pass US counterparty check across a dozen government databases now takes seconds and costs cents via API.
• SAM.gov republishes OFAC designations. A hit on both is one designation, not two.
• The same sanctioned company carries different IDs across Treasury and Commerce lists, so deduplication needs care.
• Check whether each UCC filing is still in force. Most filings on record have lapsed.
• State registries confirm existence, not ownership. They won't show a change of control.
• A US-only stack misses EU, UK and UN sanctions, beneficial ownership and adverse media.

Sources & further reading

• US Treasury, OFAC Sanctions List Search
• International Trade Administration, Consolidated Screening List
• SAM.gov, entity and exclusion search
• SEC EDGAR, full-text and company search
• GLEIF, LEI search
• Colorado Secretary of State, business and UCC search
• US Department of Labor, OFLC performance data (H-1B and PERM)
• Thoma Bravo, press release on the Ping Identity acquisition
• AI in courts: how Latin American judges are drawing the line