Proposal captured
The system records the actor, target, requested operation, claim, policy context, scope, and expected consequence before reliance.
G‑14 Proof
G‑14 proof packets are designed to show what was proposed, what evidence was checked, what decision was made, what changed, and what can be verified later.
The system records the actor, target, requested operation, claim, policy context, scope, and expected consequence before reliance.
Policy, source rights, confidence, freshness, custody, authority, and repair requirements are evaluated before action or conclusion.
Proceed, hold, repair, reject, block, or release decisions remain tied to the same proposal and evidence boundary.
The packet preserves request, evidence, decision, outcome, timing, and verifier references for review outside the UI.
Receipt forms
The artifact name changes with the workflow. The standard does not: consequential AI should leave a record that ties proposal, evidence, decision, outcome, and review boundary together.
For tool calls, deployments, writes, robot commands, workflow approvals, and other effects that need admission before execution.
Verifier-oriented packetFor agent communication, reconstruction, scientific claims, repair cycles, and reject-with-witness outcomes.
Verifier-oriented packetFor TargetLock physical-AI decisions: preserve, correct, hold, block, release, and outcome evidence.
Verifier-oriented packetRequired for an enterprise reviewer to trust the action or conclusion record.
Required for an enterprise reviewer to trust the action or conclusion record.
Required for an enterprise reviewer to trust the action or conclusion record.
Required for an enterprise reviewer to trust the action or conclusion record.
Required for an enterprise reviewer to trust the action or conclusion record.
Evaluation path
A serious proof review starts with one workflow, one policy boundary, one receipt format, and one outside reviewer who needs the record to survive inspection.