Anthropic Unveils FlashProve, an Open Source System to Prove AI Model Identity
Anthropic is working on an open source system called FlashProve that verifies which AI model produced a given output, according to a blog post published Tuesday. The system provides what researchers call provable inference — a cryptographic proof, essentially a mathematical guarantee, that an output came from a specific model. It works like a cryptographic email signature, letting users confirm that an answer really came from the model they think they are using, even against impostors attempting to disguise one model as another, including the lab that built it.
The researchers say the system could prevent an AI company from secretly swapping in a cheaper model to cut costs or a more expensive one to inflate performance results. It could also verify third-party claims, such as a developer saying a software library was scanned by a frontier cybersecurity model, or a company asserting that a business plan received approval from an appropriate model. Labs would publish a fingerprint for each released model that later gets matched to outputs and verified cryptographically.
Crucially, the fingerprints do not reveal a model's weights, the internal numbers that closed-model makers like Anthropic and OpenAI keep private, so labs can participate without giving up their most valuable assets. The system is lightweight enough that verification could run on a local device. Researchers release a technical specification and working code on GitHub under an open source license, though there are no plans to deploy FlashProve at Anthropic yet, and the team invites others to build on and improve the work.