Anthropic's new prototype proves which model wrote an AI response without exposing its weights
Anthropic released FlashProve, a research prototype that uses cryptographic methods to verify the provenance of AI responses without revealing model weights. The system aims to prove which model generated a given output,...