Styxis
Company
Knowledge

Keep person and company identity evidence separate when updating sanctions data

Collecting the latest official list and deciding whether a counterparty matches it are separate operations.

Fresh data can still produce a wrong match

A refreshed sanctions list does not establish accurate counterparty identity. Shared names, aliases and different identifier systems can still create a mistaken connection. Truthound can check the collected list's declared keys and formats; it should not make the legal identity decision.

The October additions

On October 12, 2026, the EU announced 1,569 additional individuals and entities associated with Russia's military-industrial complex, distinguishing 826 companies and 743 individuals. Electronics, semiconductors and communications suppliers are among the described sectors. Those figures define the announced list, not a finding about a particular counterparty.

Do not merge subjects by name alone

An input table should distinguish source-list version, subject type, official identifier, alias and effective date. A search may use the same name field for people and companies without applying the same confirmation rule to both. Preserve an alias's original form and relationship evidence. Do not silently replace an absent identifier with a name string.

Preserve the original basis

Keep original text apart from the internal normalised form. Otherwise a candidate created by a transformation becomes difficult to explain. Record which rule proposed the connection. Where fast screening conflicts with an opportunity to challenge mistaken identity, a review state is more informative than premature automatic confirmation.

The format boundary Truthound can check

Truthound Core Python SDK 3.3.4, operating on a prepared Polars table, provides UniqueValidator for selected-column duplicates and RegexValidator for caller-supplied string patterns. These can identify duplicate list keys or identifier-format exceptions. A pattern match is not evidence that two records describe the same person or company. The responsible owner must establish the pattern and its scope.

Keeping existing duplicate marking

An existing pandas 3.1.1 list workflow can retain DataFrame.duplicated with explicit subset and keep settings. Detecting repeated rows and resolving names remain different tasks. Compare a Truthound contract where reusable duplicate definitions are useful, without inferring an improvement in identity accuracy.

Where a shared name should stop

Keep a candidate based only on a shared name unresolved while additional identifiers are investigated. Distinguish not listed from not yet established. Otherwise incomplete research can be consumed downstream as clearance. Preserve the reason for review and the evidence needed by the next owner.

Record a correction's effect

A correction should identify which list fact or candidate connection changed. Updating current applicability and explaining a past hold have different purposes. Preserve that distinction under the responsible owner's access and retention policy rather than letting the validator decide it.

Separate screening from judgment

The useful output separates a technically checked source list from candidate relationships still needing review. List refresh, technical screening and legal judgment need different ownership. A passing input check should never silently bridge those boundaries.

Conditions

A responsible owner confirms the official list, effective information, identity rules and review process.

Alternatives

Retain pandas 3.1.1 duplicate-candidate marking separately from legal identity decisions.

Limitations

  • Format and duplicate checks do not determine name matches, ownership or control, or transaction legality.

Evidence

Event checked
Subject checked
Modified