Anthropic publishes FlashProve, a way to check which model answered

Anthropic is working on an open-source system, FlashProve, that would let someone check which model produced a given answer, TechCrunch reported Tuesday. The company describes it as provable inference: a cryptographic proof that an output came from a specific model. The researchers say it is meant to stop a lab from quietly swapping in a cheaper model to cut costs, or a more expensive one to inflate a result, and to let a user check a third party's claim that a particular model did a job.

Labs would publish a fingerprint of each model that can be matched to outputs without revealing the model's weights. Verification is designed to run on a local device. Anthropic released a technical specification and working code on GitHub under an open-source license. It said the post is intermediate research, not a finished system, and that there are no plans to deploy FlashProve at Anthropic or at any other lab.