Anthropic releases FlashProve to verify which AI model produced an answer
The open-source FlashProve prototype from Anthropic creates cryptographic proofs that an answer came from a specific model without exposing its weights.
Anthropic has released FlashProve, an open-source prototype that lets users confirm an answer came from a specific AI model without exposing that model's private weights. Weights are the trained parameters that define the model, so the design lets outside parties check an answer's source without the provider handing over the model itself.
Anthropic says the research is not yet planned for deployment.
What changed
Anthropic has released FlashProve, an open-source prototype that creates cryptographic proofs that an answer came from a specific model without exposing its weights.
Type: Research · Applies to: FlashProve
Why it matters
Companies paying for model APIs could in principle check which model actually served their requests, but the tool is a research prototype with no deployment date.
How the story developed
The Anthropic post adds the benchmark figures, the Lean formal verification of the security properties, the Apache 2.0 license and the stated limitations of the prototype.
Original source
Where we saw it
Anthropic Research, Forward Future (Matthew Berman)
AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works