Back to News
Technology
Nov 5, 2026189 views2 min read

Anthropic Releases FlashProve to Cryptographically Verify AI Model Responses

Anthropic released FlashProve, an open-source research prototype that uses cryptographic proofs to verify that an AI response came from a specific set of model weights. The tool uses zkSNARKs, a type of zero-knowledge proof. It is designed to help users confirm they are interacting with the model they think they are.

Anthropic Releases FlashProve to Cryptographically Verify AI Model Responses
Source:The Neuron

Anthropic has released FlashProve, an open-source research prototype that uses cryptographic proofs to verify that an AI response originated from a specific set of model weights. The tool was released November 5 alongside other updates to the company's Claude platform.

FlashProve uses zkSNARKs, a type of zero-knowledge proof, to generate a cryptographic certificate for each AI response. The certificate allows a user to confirm that the response came from a committed set of model weights, without revealing the weights themselves.

The tool addresses a growing concern in the AI industry: as AI systems become more capable and widely deployed, users have limited ability to verify that they are interacting with the model they think they are. A provider could, in theory, swap out a model without users knowing.

"This is about trust and verifiability," said one Anthropic researcher. "We want users to be able to confirm what they are actually talking to."

FlashProve is a research prototype, not a production tool. Anthropic said it is releasing it as open source to invite feedback from the cryptography and AI safety communities. The company acknowledged that the current implementation has significant performance limitations and is not yet practical for real-time use.

The release comes as the European Union's AI Act is pushing AI companies to implement watermarking and provenance tools. Both OpenAI and xAI are implementing text watermarking across their APIs to comply with the regulation.

Anthropic also announced beta support for message threads in the Claude Messages API, allowing server-side storage of conversation history to improve response latency. The company separately signed its first paid content licensing deal with the Associated Press, giving Claude access to verified election coverage.