Evidence and auditability

Immutable evaluation evidence

A completed Check preserves:

• Applicant and typed subject input;
• normalized provider query;
• requested datasets and promoted source versions;
• policy version and complete policy snapshot;
• result, decision, error, and timing;
• explicit Flow, Batch, Monitor, and previous-Check lineage.

Each nested matches entry preserves its stable key, material hash, score, categories, datasets, source properties, explanations, and Monitor change classification. Later policy, source, or customer workflow changes do not rewrite this evidence.

Retrieve one Check as JSON or Markdown from the same resource:

GET /v1/checks/{check_id}
Accept: application/json

GET /v1/checks/{check_id}
Accept: text/markdown

GET /v1/checks/{check_id}
Accept: application/pdf

Technical audit, not case narrative

Fidera audits execution, policy changes, Monitor runs, Batch completion, webhook configuration, and delivery attempts. Completed Check evidence is read-only.

Customer assignment, investigation notes, disposition, four-eyes approval, and case narrative belong in your system and are not copied into Fidera screening evidence.

Check representations

GET /v1/checks/{check_id} produces JSON, Markdown, or PDF according to Accept:

• JSON for machine reconciliation;
• Markdown for human-readable, diffable records;
• PDF for fixed examiner or case attachments.

Each representation includes nested match evidence and lineage. There is no export resource or separately generated report record.

Scope of claims

Evidence records what ran and how it was evaluated. It does not establish universal source coverage, precision, recall, or an accuracy claim. Dataset rights, representative corpora, persisted validation reports, and independent assurance are separate release controls.