Learn
Straight answers on authorizing the act.
Short, direct explanations of the questions buyers, builders, and regulators ask about proving who authorized a sensitive action.
How do I verify a human authorized an AI agent action?
Per-action cryptographic signatures, and why a login-time check cannot answer this.
What is cryptographic proof a human approved an action?
The four properties of a real proof, and how it differs from a log or a risk score.
The session is valid, but the action was not approved by a human.
The inheritance gap, and how to close it with execution-time authority.
Why does a passkey sign the session, not the action?
Where passwordless login ends, and where action-bound proof begins.
Payment authorization is not task authorization.
Agentic commerce authorizes the charge. The mandate for the task is the missing evidence object.
Policy gate vs signed proof
Four approaches to human approval, compared by mechanism and by the evidence each produces.
Glossary
Execution Authority Infrastructure, PSEA, execution-time authority, and the four enforcement tiers.