Service update · Anthropic
Anthropic open-sources FlashProve to verify which AI model produced a response
- Announcement time
- Review updated
Anthropic has introduced and open-sourced FlashProve, a cryptographic approach that lets developers verify whether a response came from a specified AI model. Anthropic says it is more efficient than earlier open-source methods and describes potential uses in preventing chat hijacking and checking that safety evaluations test the model actually served to users. Developers who need inference-provenance checks can explore the official write-up and code. FlashProve is not a new model and does not establish that an answer is factually correct.