Canonical transaction
Submit a factual claim and evidence, then let the protocol canonicalise the request before any model work starts.
Verification infrastructurefor AI claims and business decisions.
For agents, developers and organisations that cannot rely on an unsupported AI answer, Verified is independent claim-verification infrastructure that checks one factual claim against supplied evidence and returns a signed proof-of-claim receipt.
Protected lane
The firewall screens the submission and records the decision path so unsupported claims never blur the trust boundary.
Four-route verification
One primary and three heterogeneous judges review independently before deterministic Verified adjudication returns the public outcome.
Proof-of-claim receipt
The receipt records the evidence trail, dissent, cost and integrity state for later audit, replay and citation.
Only three public signals.
The protocol reduces complex model interaction into three public outcomes so the customer sees the result instead of a vague confidence score.
- VERIFIED
- CHALLENGED
- UNVERIFIED
Promise boundary
Verified does not promise absolute truth. It proves which verification process ran, under which policy, against which evidence, at which time.
That boundary keeps the product useful for agents and organisations that need a receipt they can inspect, share and audit later.