Example: Audit Trail

An AIEP audit trail is the complete, self-contained record of an AI response: what was asked, what was retrieved, how it was hashed, and what the validation said.

This example walks through the full chain.


Step 1 — Input

Question: How can an AI answer be verified?
Submitted: 2026-04-25T00:00:00Z

Step 2 — Evidence retrieval

Three source artefacts were retrieved:

[ev_001]  https://aiep.dev/schema/evidence.json       → fetched OK
[ev_002]  https://aiep.dev/schema/response.json       → fetched OK
[ev_003]  https://aiep.dev/.well-known/aiep/index.json → fetched OK

Step 3 — Hashing

Each artefact was hashed with SHA-256 at retrieval time:

[ev_001]  sha256: f3a2c4e0b6a7d991d3e2f5a9c8b7e6f1d0c9b8a7e6f5d4c3b2a1908172635445
[ev_002]  sha256: a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1b2
[ev_003]  sha256: 9f8e7d6c5b4a3928171615141312111009080706050403020100ffeeddccbbaa99

These hashes are the tamper-evident record. Any future change to the source will produce a different hash on replay.


Step 4 — Response generation

With the evidence rail committed, the response was generated:

Answer:
  An AI answer can be verified by binding claims to evidence artefacts,
  hashing those artefacts, and validating the response against a declared
  evidence rail.

Step 5 — Validation

The complete response was validated against the AIEP canonical schema:

Schema conformance:        passed
Hashes present:            passed  (3 of 3)
Evidence rail non-empty:   passed
Replay metadata present:   passed

Overall status:            PASSED
Generated at:              2026-04-25T00:00:00Z

Step 6 — Audit record

The full audit record in machine-readable form:

{
  "aiep_version": "2.0.0",
  "question": "How can an AI answer be verified?",
  "answer": "An AI answer can be verified by binding claims to evidence artefacts...",
  "evidence_rail": [
    {
      "artefact_id": "ev_001",
      "source_url": "https://aiep.dev/schema/evidence.json",
      "content_hash": "f3a2c4e0b6a7d991...",
      "retrieved_at": "2026-04-25T00:00:00Z",
      "validation_status": "passed"
    }
  ],
  "validation": {
    "schema_valid": true,
    "hashes_present": true,
    "evidence_rail_present": true,
    "replay_metadata_present": true,
    "status": "passed"
  },
  "generated_at": "2026-04-25T00:00:00Z"
}

This record is self-contained. An auditor can independently re-fetch each source_url, recompute the SHA-256, and compare against the recorded content_hash.


Download the full JSON

response.json →

evidence-rail.json →


Try it

Verification Playground →

View the verified response example →

GitHub →

Machine endpoint →


See also: AI Audit Trail · Verify AI Output · RAG vs AIEP · Example: Evidence Rail