Anthropic on Tuesday published a research post on FlashProve, an open-source system for verifying which model produced a given output. According to TechCrunch, it is built to provide "provable inference": a cryptographic proof, essentially a mathematical guarantee, that an output came from a specific model. Like a digital signature on an email, it is meant to hold up even when the impostor is the lab that built and runs the model.
The researchers say the system could stop an AI company from quietly swapping in a cheaper model to cut costs, or a more expensive one to inflate results. It could also verify third-party claims, for example that a software library was scanned for vulnerabilities by a frontier cybersecurity model.
To work, AI labs must publish a "fingerprint" for each model they release. The fingerprint does not reveal the model's weights, so makers of closed models could use it too. The researchers released a technical specification and working code on GitHub under an open-source license, but according to TechCrunch the research is ongoing and there are no plans to deploy the system at Anthropic or any other lab. "This research post is meant to share intermediate progress," the researchers wrote.