Anthropic is developing FlashProve — a system for cryptographically verifying which AI model produced an answer
Anthropic has unveiled a research project called FlashProve: an open “provable inference” system that makes it possible to cryptographically verify that an answer was genuinely generated by a specific model, including protection against model substitution by the lab itself. A specification and code have been published on GitHub, but no one plans to deploy the system yet.
Aziz Karimov