Anthropic’s FlashProve uses zero-knowledge proofs to verify large AI model outputs

FlashProve's cryptographic verification could enhance AI model trustworthiness, but high computational costs hinder widespread adoption.