Anthropic Open-Sources FlashProve to Verify AI Model Identities
Anthropic is developing and open-sourcing FlashProve, a tool that uses cryptographic proofs to verify if an AI output was generated by a specific model. It helps identify model substitution and verify third-party performance claims. Labs must publicly release model fingerprints for future matching—these do not disclose model weights, and verification can occur locally. Anthropic has already published technical specifications and code. Research continues, with no current deployment plans.
Anthropic Open-Sources FlashProve to Verify AI Model Identity
Anthropic is developing and open-sourcing FlashProve, a system that uses cryptographic proofs to verify whether a given AI output was generated by a specific model. Researchers say the system can help users identify model substitutions and also verify third-party claims about a model's performance on specific tasks.
Laboratories must publicly release a fingerprint for their models to enable subsequent output matching. This fingerprint does not reveal the model's weights, and verification can be performed on local devices. Anthropic has already published the technical specifications and code. Research is ongoing, and there are currently no plans for deployment.
Claude | GitHub | TechCrunch