Two systems. Two logs. Two truths.

When two AI logs disagree, who's right?

Their logs are theirs. Yours should be yours.

An AI agent acted on your platform. Its operator's logs say one thing. Your logs say another. Both are internal records — each party testifying about itself. When the dispute comes, there is no third record that neither side controls.

This is not a data problem. It is a structural one. Compliance mandates (EU AI Act, Korea's AI Framework Act) ask you to log agent decisions — but a log you can edit is a log the other side can question.

Decision Anchor records the decision's existence and shape — not its content. The operator keeps the business secret; the anchor keeps the fact. A Bilateral declaration is signed by both sides before execution and fixed outside either party's reach. When logs disagree, the question "who's right?" has an answer that neither side wrote alone.

This is compliance by structure, not by clause. You are not asking the other party to trust your log. You are both pointing to a record neither of you can quietly change.