· Yair Knijn
The sanctioned UBO no name-match ever finds: how the OFAC 50% rule hides behind a holding company
An underwriting director runs the named insured through the screening tool, gets a clean response, and binds the risk. The legal entity on the application is real, it has a trade register number, and it does not appear on the SDN list. So the file shows no hit. The director has answered the wrong question, because the entity on the application is almost never the party OFAC cares about.
The party OFAC cares about sits two layers up the ownership chain, and a name screen of the front company will never reach it.
Why screening the named entity is the wrong unit of analysis
A name screen compares one string against a list of strings. It is a matching exercise, and it answers exactly one thing: is this specific name, or something fuzzily close to it, on a watchlist today. That is useful, but it is not what blocks a risk. OFAC blocks an entity based on who owns it, not on whether its name happens to be enumerated. The named insured can be perfectly clean as a string and still be a blocked person as a matter of law.
So the unit of analysis is wrong. You screened a company. The exposure lives in the natural persons behind it, and those persons were never in the query.
The OFAC 50% rule and the trap of aggregated, indirect ownership
OFAC's 50 Percent Rule states that any entity owned 50 percent or more, in the aggregate, by one or more blocked persons is itself blocked, whether or not it appears on the SDN list. Two words there do the damage. Aggregate means you sum the stakes: two SDNs holding 25 percent each produce a blocked entity, and neither stake alone trips a threshold a junior reviewer would flag. Indirect means ownership held through intermediate entities counts in full, so a clean-named holding company that is itself 50-percent-plus SDN-owned passes its blocked status straight down to everything it controls.
OFAC reinforced this in its March 31, 2026 guidance on sham transactions and sanctions evasion, which treats the 50 percent rule as a floor for diligence rather than a ceiling. The regulator's position is that you are expected to look through structure, not stop at the first clean name. Picture two designated individuals, each holding shares in a Cyprus intermediate, which holds 51 percent of the operating company on your application. Every name on your slip is clean. The risk is blocked, and you wrote it.
Ownership layers, nominees, and the parties your name screen never sees
Real structures are built to defeat exactly the check most insurers run. The operating company is owned by a holding company, owned by another holding company in a second jurisdiction, with a nominee director of record and the beneficial owner appearing nowhere on the application. None of these intermediate parties are on your slip, so none of them are in your screening query, so the natural persons who would actually trip the SDN list are invisible to the process end to end.
- The named insured returns clean because it is a string the tool can match and it is not listed.
- The intermediate holding companies are never screened because nobody asked who owns the named insured.
- The natural-person owners, the only parties the 50 percent rule turns on, were never collected, so they were never queried.
Demanding UBO declarations and screening every owner, not just the front company
The fix is not a better fuzzy-match algorithm. It is a different input. Require a beneficial ownership declaration as a binding condition, down to the natural persons, with percentages at each layer so you can do the aggregation OFAC actually requires. Then screen every one of those persons and every intermediate entity, not only the name on the slip. When ownership is opaque or the declaration is refused, that is itself an underwriting signal, and the honest answer is to decline rather than to record a clean name screen as if it were diligence.
A clean entity match is meaningless if you never resolved the people who own it. InsureGuardAI builds the screen around the structure rather than the string: each workspace captures the ownership chain, screens every owner and intermediate entity against the SDN list, aggregates indirect stakes to the 50 percent threshold, and keeps the raw match response as evidence so the file answers the ownership question before a regulator thinks to ask it. See how it works.