Tamper-Evident Audit Logs: How a Hash Chain Proves Nothing Was Altered
The point of an audit log is not that it records events. It is that you can tell when someone changed the record.
Most systems keep a log. Few keep a log you can trust after the fact. The difference is whether a person with access could quietly edit a past entry and leave no trace. A hash chain closes that gap: it does not stop edits, it makes them detectable.
The mechanism, plainly
Each event stores a cryptographic hash of its own contents plus the hash of the previous event. Because every link folds in the one before it, the entries form a chain. Change any past event — even a single character — and its hash changes, which breaks the link to the next event, and every event after it. Verification is just walking the chain and checking that each recorded previous-hash still matches. One mismatch tells you exactly where tampering began.
A hash chain does not prevent tampering. It makes tampering impossible to hide.
Why it matters for AI agents
When an agent takes actions on your behalf, the log is the accountability. "The agent renamed a profile, ran a check, sent an alert" is only meaningful if you can trust the record was not edited after the fact — by a buggy process, a bad actor, or the agent itself. A tamper-evident chain turns the log from a convenience into evidence.
What it is not
A hash chain is not encryption and not a blockchain. It needs no network, no consensus, no coin. It is a small, local discipline: hash the event, include the previous hash, store both. It also does not, by itself, prove when something happened — pair it with trusted timestamps if ordering in real time matters. Used honestly, it should be labeled for what it is, not oversold.
The takeaway
If a system's audit log can be edited silently, its promises are only as good as everyone's honesty. Chain the events, and the record starts defending itself. That is the quiet backbone under a trustworthy agent harness.