Anthropic Uses StarkWare Cryptography to Identify Models Behind Answers

StarkWare said Anthropic is using cryptography co-invented by the firm’s founders to identify which AI model produced a specific answer, applying the technique to model validation.

The announced use connects an answer with the model that generated it. Starknet amplified StarkWare’s announcement on X.

StarkWare co-founder Eli Ben-Sasson has separately said ZK STARK and Starknet could contribute to AI safety. StarkWare’s Applied AI & Verifiable Intelligence lead Abdel Bakhta said Oct. 23 he was leaving the company to focus on AI safety. Ben-Sasson welcomed Bakhta’s decision.