FreeLLMAnthropic releases open‑source FlashProve to cryptographically verify AI model outputs
Anthropic's FlashProve lets anyone prove which model generated a response without exposing model weights, using zero‑knowledge proofs.
Anthropic announced a prototype called FlashProve that can produce a cryptographic proof linking an AI output to a specific model.
The system relies on labs publishing a non‑revealing fingerprint for each model, enabling verification even if the model's weights stay private.
FlashProve runs on a full‑size Llama 4 Maverick model with a 1,024‑token context window using a single NVIDIA H100 GPU, achieving about 110 tokens per second.
Verification can be performed on a regular CPU in seconds, and the code, technical specifications, and examples are available under an Apache 2.0 license on GitHub.
Anthropic emphasizes that the prototype is not slated for deployment yet and is shared for the community to build upon.
O que isso significa para você
For AI users, FlashProve offers a way to confirm that a response truly comes from a claimed model, protecting against hidden model swaps. Developers of free or locally run models can adopt the open‑source tools to add provenance guarantees without exposing proprietary weights.
Fontes
- Anthropic is working on a system to prove AI models' identity, TechCrunch, 3 Nov 2026
- GitHub - anthropics/FlashProve: A research prototype for zero-knowledge proofs of AI inference. Not maintained. · GitHub, github.com (official), 3 Nov 2026
Escrito automaticamente a partir das fontes acima, depois verificado pelo nosso código: todo número e nome deve aparecer nelas, e nenhuma frase pode copiar uma fonte. Encontrou um erro? Reporte‑o.
Qual IA você pode usar grátis hoje?
Todo modelo de IA gratuito, classificado por votos humanos cegos, com a memória e velocidade que ele precisa na sua própria GPU.