- Human oversight evidence in the EU AI Act decision record
For the compliance officer who has to show, for one named high-risk system and one named date range, which AI decisions a human overrode.
- Clinical AI decision documentation without ingesting diagnosis codes
Auditrek is built to log what a clinical decision support model output, how confident it was and whether the clinician overrode it, as SHA-256 hash-chained records, with an ingestion gate that rejects recognisable raw sensitive fields, including diagnosis code patterns, before anything is written.
- Algorithmic hiring audit trail: one hash-chained record per candidate decision
Auditrek is built to capture a hash-chained, append-only decision record for every output your hiring screening model produces, with candidate identifiers pseudonymised at ingestion and chain verification recomputed server-side.
- Logging AI decisions without touching the inference path
A page for the ML engineer who has been handed the audit-trail problem: Auditrek takes inference events over REST or webhook after the prediction has returned, so decision logging never enters the serving path.
- Underwriting model audit trail: decision records your actuaries and auditors can both read
Auditrek records each underwriting or pricing inference as a hash-chained decision record, then generates Article 12 and 13 field reports by date range.
- Credit scoring: an append-only record behind every automated decision
A credit scoring model audit trail is a per-decision record of which model version scored which application, what output came back, when it happened and whether a person overrode it.