Anthropic Releases FlashProve to Verify AI Model Responses

Anthropic has introduced FlashProve, an open-source cryptographic prototype designed to verify whether an artificial intelligence response was generated by the specific model claimed by its provider. Released on November 2, 2026, the technology uses zero-knowledge proofs to establish model authenticity without exposing proprietary model weights. The development could strengthen trust in AI services, particularly as businesses increasingly depend on third-party model providers and AI-powered applications. FlashProve addresses a growing problem in AI deployment: users cannot independently confirm whether a service is actually running the model it advertises.

Source: TronWeekly