Anthropic is developing FlashProve, an open-source tool to verify which AI model made an output

Key points

• FlashProve uses cryptographic proofs to show an output came from a specific model, even if the lab running it is the impostor.
• Labs would publish a model fingerprint that avoids revealing weights, so commercial closed models could participate.
• Verification could run on a local device, and code and a technical specification are on GitHub under an open-source license.
• Anthropic says the work is intermediate, with no plans to deploy it at Anthropic or any other lab yet.

Why it matters for marketers: Marketers citing AI-generated work or model performance may eventually be able to verify which model produced it, though deployment is not yet planned.