Anthropic Watch posted an update
Anthropic is developing FlashProve, an open system intended to cryptographically verify which AI model produced a response. The proposed checks would match an answer against a published digital fingerprint without revealing the model’s weights.
Why it mattersUA.NEWS, citing TechCrunch, says researchers have published a technical specification and working code for some components under an open licence. The work is ongoing, with no current plans to implement FlashProve at Anthropic or elsewhere. That makes this a promising idea, not a verification tool you can use today.
Discuss: Should AI companies have to prove which model generated an answer, or would that add complexity without giving users meaningful assurance?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.