Anthropic develops FlashProve system to verify AI model identity

Anthropic is developing FlashProve, an open-source system that uses cryptographic proofs to verify whether a given AI output came from a specified model. The approach requires model providers to publish a model "fingerprint" that users can match against outputs without exposing model weights. Researchers say the tool could prevent providers from quietly swapping models and allow verification of third-party claims about which model they use. Anthropic has released the technical specification and some system code, but the research is ongoing and there are no current plans to deploy FlashProve at Anthropic or other labs. What to watch next is whether providers begin publishing fingerprints and whether the system moves from research code toward production use.