Sanctions Desk gathers and checks the evidence for every sanctions-screening alert and recommends an outcome; a person with the right entitlement approves it. One alert, start to finish · The problem it works on

Replay of recorded runs. Nothing runs live here. Every value below is read from files committed in the repository. To run the desk yourself, follow the README quick start.

What happened: The decision step proposed a close once; the critic approved it; the policy guard let it stand. Outcome: Close as a false positive (CLOSE_FALSE_POSITIVE), as the label expects.

Outcomes: CLOSE_FALSE_POSITIVE = close as a false positive · ESCALATE_L2 = escalate to level 2 · NEED_MORE_INFO = ask for more information · RECOMMEND_BLOCK = recommend a block (two people must confirm). Raise-only: the policy guard can make an outcome more cautious, never less.

A close: a spelling variant of a listed name with a different date of birth and nationality; the critic approved it, the challenger found no credible argument against it, and the guard let it stand for a person to confirm.

Outcome

Desk recommendation

Close as a false positive (CLOSE_FALSE_POSITIVE)

Label in the case file

Close as a false positive (CLOSE_FALSE_POSITIVE) (eval/cases/exam.json, case EX-004)

Compared with the label

matches the label

Code's first finding

Close as a false positive (CLOSE_FALSE_POSITIVE)

Critic

1 round(s), final: approved

Model calls

4 (1 local, 3 cloud); 4 with a signed receipt, all verified

Run

sys2b-routed-20261023T195504Z, policy version 6, 36 steps in 73.1 s

Case note from the case file: Similar name, but the date of birth conflicts and the nationality differs; no document matches.

Why the close stood

No check stopped this close. It is still only a recommendation: a person confirms or changes it. What each check said:

• critic (step 29): approved
• challenger: no credible argument: While the name similarity is high (0.965) and 'Endrew' could be a transliteration variant of 'Andrew', the customer's date of birth (1956-12-27) conflicts with the listed party's date of birth (1969-09-22) at day precision, and the customer's nationality (CH) does not match the
• policy guard (raise-only rules in code): CLOSE_FALSE_POSITIVE stands

Steps in order

Each line is one event from the run's trace (trace.jsonl.gz). The model is the one that served the call, as the signed receipt records it.

Model calls and receipts

Each signature was checked when this page was built, with the same code as npm run evidence:verify (Ed25519 over the receipt body in receipts.jsonl.gz); the build stops if one does not verify, is missing, or names a different model from the trace. The Nightshift runtime's receipt key is pinned in config/default.json, so signature valid, key pinned shows the receipt was not changed and that this runtime signed it; a receipt signed by any other key stops the build.

The whole run passes the same check: 305 receipts with valid signatures, 305 model calls in the trace that name a receipt, none missing and none naming a different model from its receipt.

Signing key fingerprint (first 16 hex of SHA-256): 1835d1eab9211f97.

Zetaris queries

This run's export keeps the table, the row count and the Zetaris query id of each read, not the statement text; the statements are not reconstructed here.

Check it yourself

Clone the repository and run npm run evidence:verify -- eval/runs/sys2b-routed-20261023T195504Z. What the export left out: tokens, local paths, hostnames and addresses other than the local machine, usernames, e-mail addresses and secret environment values were removed from results.json and trace.jsonl.gz; receipts.jsonl.gz is untouched.