Anthropic Prototype Verifies Which AI Model Generated a Response
🤖 AI-generated content — The title and summary were produced automatically by artificial intelligence, without human editorial review.
Anthropic’s open-source FlashProve prototype uses zkSNARK proofs to verify an AI response came from a claimed model without revealing its weights. For the 400-billion-parameter Llama 4 Maverick, it produces 19MB proofs at about 110 tokens per second on one H100 GPU; CPU verification takes 17 seconds, with 70–95 times normal inference overhead.
Continue in the app — vote & join in ➔Source: The Decoder (AI) · via ahirlevel.hu