KPATH

What makes an AI agent audit log evidence?

A log becomes evidence when a third party can verify it without trusting whoever produced it. KPATH records every decision in a signed, tamper-evident record that streams to your SIEM and is ready for write-once retention, and an auditor can prove offline that the record has not been altered without trusting KPATH.

A log you control is not evidence.

The problem with a log

Almost every system writes a log. Almost none of those logs would survive an adversarial reading. The person who runs the system can edit it, and the person asking the question has to take the operator’s word that they did not. When the question is “did this agent move money after it was told to stop”, the operator’s word is not what a regulator, an insurer or a counterparty wants.

That is the standard KPATH sets for the Prove step of the Verify-Enforce-Stop-Prove test: a record a third party can check without trusting the party that produced it, including without trusting KPATH.

What KPATH records

Every call that passes through KPATH lands in the record, allowed or refused. So does every approval, captured as it happened, and every guardrail verdict. Stopping a chain writes a containment proof showing that nothing in the chain was authorised afterwards.

The record holds who acted and what was decided: agent identity, owner, risk tier, verified principal where one exists, target, action, decision, approval and outcome. It never holds a secret or a message body, because it is built from the request envelope. The record cannot leak what KPATH never saw.

What the auditor gets

Property What it gives the auditor
Tamper-evident Any edit, deletion or insertion shows up
Authentic The record was produced where the decision was made, not reconstructed later
Independently checkable It verifies without KPATH’s say-so
Durable Held on write-once storage, a retained record cannot be rewritten
Verifiable offline A third party proves the record on their own machine, with no KPATH console

Together these mean an auditor can establish that the record has not been altered without taking our word for it. That is the difference between a log and evidence.

Verifying without trusting KPATH

Verification is deliberately outside the product. A third party can take an exported record and satisfy themselves that every entry is authentic and that nothing has been added, altered or removed, on their own machine, without calling KPATH and without KPATH’s console. A third party can also generate a known sequence of calls, stops and approvals and confirm that the record reflects exactly that sequence.

KPATH’s own console is not the source of trust. If KPATH disappeared tomorrow, the exported record would still prove what happened.

Streaming to your SIEM

Evidence you cannot find is not much use either. Every record streams to your SIEM as it happens, so it lands where your auditors and your security operations team already read, ready for write-once retention.

Because the record contains no payload and no secrets, streaming it outside the enforcement boundary creates no new exposure.

Answering the auditor

Supervisors hold an agent that moves data or money to the same standard as the person who used to do the job. Their questions arrive in the same order every time.

Question What the record shows
Who is this agent? The verified identity, its named owner and its risk tier on every call
What is it allowed to do? The policy decision on every call, including refusals
What did it actually do? Every call, approval and outcome, in order
How did you stop it? The stop, and a containment proof that nothing was authorised afterwards

Why it is built this way

An audit trail inside the agent, or inside the framework the agent runs on, is written by the thing under investigation. A log in a console the vendor controls is written by a party with an interest. KPATH sits outside the agent, records at the point the decision is enforced, and then hands you the means to check the record without KPATH. A log you control is not evidence. A record anyone can verify is.

FAQ

Common questions

What is in the audit record?

Who acted and what was decided: the verified agent identity, its owner and risk tier, the principal where one exists, the target, the action, the policy decision, any approval and the outcome. Never a secret or a message body.

Can our auditor verify the log without KPATH?

Yes. The record is signed and tamper-evident, and it can be verified offline. The auditor does not need KPATH's console, or KPATH's word, to confirm that the record has not been altered.

Where does the record go?

It streams to your SIEM as it happens, and it is ready for long-term retention on write-once storage.

Does the audit log contain our data?

No. The record is built from the request envelope, so it holds no secret and no message body. That is what lets the record leave your boundary for a SIEM, a regulator or a counterparty.

Published Updated

The first step

Start in monitor mode. See everything before you enforce anything.

Deploy in monitor mode: observe only, enforce no policy, rewrite no agents. Flip to enforce by repointing egress. See the agents already running in your estate before you decide what to enforce.

Book a monitor-mode pilot See the platform