Cryptographically verifiable proofs that specific disclosures are absent from AI model documentation. Pedersen commitments on secp256k1, temporally anchored, mapped to 11 regulatory jurisdictions.
A TPA is a cryptographic proof that at a specific point in time, an AI model's documentation did not contain certain required disclosures.
Each observation is committed using Pedersen commitments on the secp256k1 curve, anchored in a Merkle tree, and chained temporally so no past observation can be altered.
Detection method: structured HuggingFace API fields (cardData) — explicit metadata keys only. A NEC is absent if the corresponding field is not declared. This differs from the Compliance Ranking which uses keyword analysis of card text.
Absences are mapped to regulatory requirements across EU AI Act, GDPR, US DMCA, China AI Provisions, CCPA, ISO 42001, OECD, UK CDPA, Japan Copyright Act, UK AI, and US FTC.
Full cryptographic verification, jurisdiction articles, and temporal pressure analysis available in Crovia PRO.