Rescreening is a race between a new designation and the next payment. The agent tracks each list update, confirms the screening tool has loaded it, triggers rescreening of the customer base, and prepares every new hit for review with a side-by-side comparison. Analysts decide each match, and the sanctions officer takes any freezing and reporting action.
What is rescreening, and who owns it?
Sanctions lists change often: new designations, amendments to identifiers, delistings. Firms must screen their existing customers against the updated lists promptly, not only at onboarding. FATF Recommendations on targeted financial sanctions expect assets of designated persons to be frozen without delay, which means a firm that rescreens slowly can be holding or moving funds for a designated person without knowing it. FATF targeted financial sanctions and AI agents covers the standard in more depth.
The sanctions officer owns rescreening, with technology running the screening tool and analysts reviewing hits. The weak points are timing and confirmation: whether the update was loaded, whether rescreening ran across every customer population, and whether hits were reviewed fast enough. Whether an AI agent can rescreen customers when sanctions lists change sets out the boundary.
What are the steps, and who acts at each?
- Agent, depth 1: tracks list publications. From the authorities' published lists and the tool's list provider, recording each update's publication time and contents. This mirrors how agents track regulatory updates.
- Agent, depth 1: confirms the update was loaded. Compares the list version in the screening tool with the published version. A gap goes straight to the sanctions officer.
- Agent, depth 3, within bounds: triggers rescreening of all in-scope customer populations using the tool's approved rescreening job. It cannot change thresholds or scope.
- Agent, depth 1: confirms completion. Every population rescreened, with record counts matching the customer base.
- Agent, depth 2: prepares new hits. For each, a comparison of identifiers, a summary of the customer's activity and any pending payments.
- Person, analyst: decides each hit.
- Person, sanctions officer: acts on true matches. Freezing, blocking, and reporting to the relevant authority.
- Agent, depth 1: reports timing. Publication to load, load to rescreen complete, rescreen to hits reviewed.
| Step | The agent does | Depth | Stays with a person |
|---|---|---|---|
| Track | Records each list update | 1 | None |
| Load check | Compares tool and published versions | 1 | Fixing gaps |
| Rescreen | Triggers the approved job | 3 | Defining the job |
| Completion | Checks counts by population | 1 | Investigating shortfalls |
| Prepare hits | Drafts comparisons | 2 | Deciding each hit |
| Act | Nothing | None | Freezing and reporting |
| Timing | Reports each interval | 1 | Acting on delays |
Which controls apply, and what evidence does each leave?
Timing recorded at each stage. Evidence: a timeline per update, which is what a supervisor will ask for after a designation affecting your customers.
Completeness checked by counts. Rescreening that silently misses a customer population, such as customers on a legacy system or beneficial owners of business customers, is a common failure. Evidence: counts per population matched to the customer base.
No disposition capability. The agent prepares; analysts decide. Evidence: the agent's capability list. The reasons are set out in whether an AI agent can clear a sanctions alert.
Pending payments flagged. A hit on a customer with an outgoing payment queued needs attention first. Evidence: hits prioritised by exposure, with the order recorded.
Delistings handled carefully. A delisting may allow a frozen relationship to resume, but only after a person confirms it. Evidence: decisions on delistings by the sanctions officer.
What should you measure?
- Time from publication to list loaded in the tool.
- Time from load to rescreening complete.
- Time from rescreening to all new hits reviewed.
- Populations missed in any rescreening run.
- True matches found by rescreening, and the exposure at the time.
Where does it usually go wrong?
Relying on the provider without checking. List providers are usually fast, but not always. The agent's load check catches delays.
Related parties missed. Directors, beneficial owners, guarantors and authorised signatories must be rescreened as well as account holders. Check that the job covers them.
Hit surges. A large update produces many hits at once. Prioritise by exposure and pending payments, and bring in more analysts; do not let the agent decide hits to clear the queue.
Policy not updated after a regime change. A new sanctions regime or a change in scope may need a policy update, not just a new list. Route it through drafting policy updates for approval.
Ownership-based restrictions. Entities owned or controlled by designated persons may be restricted even when not listed. Rescreening names alone will not find them; ownership data must be checked too.
Amendments treated as noise. An update that adds a new alias or passport number to an existing entry can turn an old false positive into a true match. Rescreening must cover amendments, not only new designations, and earlier dispositions against the amended entry should be re-opened for review.
The overall stance on compliance agents is in AI compliance agents.
Where Pryme fits
In Pryme, triggering the approved rescreening job is a single depth 3 capability, bounded to that job, and the agent cannot change thresholds or scope. Deciding hits and freezing funds are outside its role scope; attempts are refused and filed. Pending payments reach the agent from the ledger as sealed facts, so exposure is ranked on records.
Questions people ask
Can the agent freeze an account when it finds a likely match?
No. Freezing is a legal action with consequences, decided by the sanctions officer on a confirmed match. The agent prioritises the hit and flags any pending payment so a person can act quickly.
Does rescreening need to run on every list update?
On every update that affects the lists you screen against. The agent's tracking makes this practical rather than relying on a fixed schedule.
What about customers who have closed their accounts?
If you still hold their funds or are processing final payments, they remain in scope. The rescreening job should include them until the relationship has fully ended.