Definition
An emerging open standard for cryptographic, publicly checkable evidence that an AI agent stayed inside the limits it was given. It aims to replace vague vendor promises with verifiable proof of what each agent actually did, and maps to NIST, ISO and EU AI Act frameworks. The v1.0 working draft was released for public comment in September 2026.
Why it matters
For buyers and regulators, verifiable proof is the difference between trusting a claim and being able to check it — expect procurement teams to start demanding it.