Anthropic develops open source system to verify artificial intelligence model identities

Anthropic is developing an open source system named FlashProve to verify the identity of artificial intelligence models, according to TechCrunch. TechCrunch reported on Tuesday that the system is designed to allow users to confirm whether an output was produced by the specific model it claims to originate from. [1]

The FlashProve system relies on cryptographic proofs, which serve as mathematical guarantees of provable inference, TechCrunch reported. Similar to cryptographic email signatures, the software is designed to protect users against impostors attempting to disguise one model as another. [1]

Anthropic operates under a governance structure that includes a Board of Directors and an independent Long-Term Benefit Trust. In 2022, the company introduced Constitutional AI, a training method that uses explicit natural-language principles to train models such as its Claude series. [3]

Large language models and reasoning models developed by companies such as OpenAI and Anthropic have increasingly contributed to research-level mathematical proofs. Anthropic disclosed details about the FlashProve project in a blog post published on Tuesday, TechCrunch reported. [1][2]