Anthropic releases FlashProve research on provable inference and open-sources the code

Original title: New Anthropic research: progress on provable inference for large language models.

Anthropic introduces FlashProve, built on established cryptographic techniques, allowing developers to verify that a given response came from a given AI model, substantially more efficient than prior open-source methods, with the code open-sourced. Applications include making it harder for attackers to hijack a chat and letting safety evaluators confirm the tested model matches the one served to customers; the team sees further progress in provable inference as important for trust in AI deployment.

The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.