Anthropic crafts system to prove AI model's identity
anthropic develops flashprove to verify ai model outputs, preventing model impersonation and ensuring transparency. this affects ai labs, developers, and users relying on ai outputs.
anthropic has announced an open-source system called flashprove, designed to verify which ai model produced a given output. flashprove provides provable inference, a cryptographic proof that ensures a specific output came from a specific model. this system is meant to prevent model impersonation, even by the lab that built and runs the model. flashprove requires ai labs to publish model fingerprints that can be matched to specific outputs and verified cryptographically, without revealing the model's internal weights. the system is lightweight and can be used with commercial models. researchers have released a detailed technical specification and working code for flashprove on github under an open-source license.