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 recommended NEED_MORE_INFO; the critic sent it back once and then approved it; the policy guard let it stand. Outcome: Ask for more information (NEED_MORE_INFO), 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 data-quality rule blocks a close: the customer has no date of birth on record, so the Zetaris rule dob_present_for_individuals fails, the policy guard blocks any close, and the case goes to a person as need more information.

Outcome

Desk recommendation

Ask for more information (NEED_MORE_INFO)

Label in the case file

Ask for more information (NEED_MORE_INFO); also acceptable: ESCALATE_L2 (eval/cases/exam.json, case EX-030)

Compared with the label

matches the label

Code's first finding

Ask for more information (NEED_MORE_INFO)

Critic

2 round(s), final: approved

Model calls

5 (1 local, 4 cloud); 5 with a signed receipt, all verified

Run

sys2b-routed-20261023T195504Z, policy version 6, 40 steps in 32.4 s

Case note from the case file: The match can be neither confirmed nor ruled out; the data-quality rule fails.

What the reviewer does next: The sealed record holds no human decision yet, so nothing has been prepared. The reviewer approves Ask for more information (NEED_MORE_INFO) or overrides it with a reason. On approval the desk prepares a request-for-information draft, which a person must approve before it is used for ALR-0030. Nothing is sent.

Checks along the way

No close was proposed in this run. What each check said:

• critic (step 29): rejected: cited fact does not hold: unknown table "data_quality". Cite only rows of these tables: account (key account_id), alert (key alert_id), beneficial_owner (key ownership_id), business_definition (key definition_key), customer (key customer_id), customer_alias (key alias_id), customer_identifier (key i
• critic (step 37): approved
• policy guard (raise-only rules in code): NEED_MORE_INFO stands. Checks that raised or flagged: data_quality: block (customer record fails dob_present_for_individuals)

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

The statements are exactly as the sealed decision record of this run keeps them (output_examples/need-more-info-critic-sent-back/decision-record.json; its hash matches the one the trace logged when it was sealed). The record keeps each statement and a hash of each row returned, not the rows themselves.

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.