Anthropic releases FlashProve for verifying AI model outputs
Original title: Anthropic's new prototype proves which model wrote an AI response without exposing its weights
FlashProve uses zkSNARKs to prove model authorship without exposing weights, limited to 1024 token contexts.
Anthropic released FlashProve, a research prototype that uses cryptography to verify AI responses from the claimed model without exposing weights. The system has limitations including support only for 1024 token contexts and deterministic outputs. Regulators and companies could use it to check approved models and detect tampering.
- Proof generation runs at 110 tokens per second on H100 GPU for Llama 4 Maverick
- Each proof is 19 MB and verifiable in 17 seconds on CPU
- Prototype is open-source under Apache 2.0 license
- Does not require trusted hardware unlike prior approaches
- Vendors
- Anthropic
- Products
- FlashProve
- AI models
- Llama 4 Maverick
From the original article
2 min read at the-decoder.com
Anthropic has released FlashProve, a research prototype that uses cryptography to verify whether an AI response actually came from the claimed model. The system still has significant limitations. The article Anthropic's new prototype proves which model wrote an AI response without exposing its weights appeared first on The Decoder.
By Maximilian Schreiner · the-decoder.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.