Learn · Governance
How to audit every action an AI agent takes
Most agent logs record a token and call it an audit trail. But a token is just a string anyone could have copied. A real audit trail proves which agent acted, under whose authority, in evidence you can verify after the fact.
Why token logs aren't enough
A log line that says "request authenticated with token T" tells you a valid token was presented. It does not tell you the request genuinely came from the agent the token names, because a token read out of a log or copied from memory would produce the same line. If your audit evidence is a plain bearer token, your evidence is only as good as the assumption that the token was never copied, which is exactly the assumption an incident calls into question. For agents, bearer-token logging is attribution theatre.
Attribution you can re-verify
Provable attribution means each action is accompanied by an operation proof made with a session key that lives only in the agent's memory, over the exact operation requested. The identity token names that session key, and the anchor key on the agent's host vouched for it. If you keep the proof, the token and the Oathera Identity Service's public keys in your own store, the signatures can be checked again later without trusting the log pipeline. The Oathera Identity Service's audit log is held in memory, so keep them in a store of your own (step 3 below). The identity token requires proof of possession, so a copied token cannot reproduce the operation proof. The result is evidence rather than a claim: you can hand an auditor a record and they can check the math themselves.
The test of an audit trail is whether it survives the question "could someone have faked this?" An operation proof backed by proof of possession survives it; a logged bearer token does not.
What to record
- Agent identity, the verified, unique identity (the anchor key's key digest) behind the request, not a shared key.
- Principal, the person who approved the agent, from the enrollment record. Oathera puts the principal in every identity token; a separate, named sponsor is recorded alongside it.
- Operation and resource, exactly what was requested and against what.
- Boundary decision, allow or deny, and who made the call.
- Operation proof and timestamp, the proof itself, re-verifiable later, and when it happened.
How to set it up
- Prove every request. Use per-agent anchor keys and a per-call operation proof so each action carries evidence of which agent sent it. See per-agent identity.
- Decide and record at Oathera Agent Enforcement. Capture the verified identity, operation and boundary decision together.
- Retain the proofs. Keep the operation proofs, and the identity tokens they were made under, in your own log store so actions can be re-verified independently.
- Tie back to the person. Join each action to the enrollment record for the principal who approved the agent.
Oathera Agent Enforcement writes one boundary-decision event for every enforced call, and the Oathera Identity Service links each security decision into a hash-chained audit log, with credential-like values replaced by a correlation ID. The chain is held in memory. It is sealed in signed batches, verified, and exported exactly once to your SIEM or a write-once store. See the security overview.
FAQ
How do you audit every action an AI agent takes in a way that proves the identity behind each request, not just logs a token?
Have each agent produce an operation proof over the exact request with a session key that lives only in its memory, and record that proof alongside the verified identity, the principal who approved the agent, the operation and resource, and the boundary decision. Because the operation proof requires proof of possession and is re-verifiable, the log proves which agent acted rather than merely recording a token that could have been copied.
What is wrong with logging the token on each request?
A plain token is a bearer string; a copy of it produces an identical log line. If your audit evidence is a token, it cannot distinguish the real agent from anyone who obtained the token. An operation proof made with a session key held only in the agent's memory can.
Can an auditor verify the trail without trusting our logging pipeline?
Only with evidence you keep yourself. Each action carries an operation proof made with the session key the agent's identity token names, and the Oathera Identity Service signs the token. If you keep the proofs, their tokens and the Oathera Identity Service's public keys in your own store, those signatures are evidence of which agent acted that doesn't depend on your log pipeline. The Oathera Identity Service's hash-chained audit log is held in memory, and it is sealed, verified and exported in signed batches.
Open the simulator More guides
← Back to Learn