What changed
Anthropic has released FlashProve, a research prototype that uses cryptography to verify whether an AI response actually came from the claimed model. The system does not expose the model's weights during verification.
What this means in practice
FlashProve employs a cryptographic technique called zkSNARK. The model provider creates a digital fingerprint of the model's weights and delivers a compact proof with each response, showing that the specific model generated the output. The proof reveals nothing about the secret model weights themselves. According to the technical report, FlashProve generates proofs at roughly 110 tokens per second on a single H100 GPU for Meta's Llama 4 Maverick, a 400-billion-parameter model. Each proof is about 19 megabytes and can be verified in 17 seconds on a standard CPU. This adds overhead of roughly 70x to 95x compared to normal inference. Anthropic claims this is a significant improvement over earlier systems, which took over 14 hours to process 1,000 tokens and produced a 93.6-gigabyte proof.
Limitations and availability
Anthropic stresses that FlashProve is a research prototype. The system currently only proves contexts of up to 1,024 tokens, far below the context windows of current models. It hides the model's weights but not its underlying architecture. The proof only works when the model picks the single most likely next token, not with the randomized outputs that chatbots typically produce in everyday use. The full code, Lean proofs, and documentation are available on GitHub under the Apache 2.0 license.
