Google DeepMind says AlphaProof Code formally verified Signal’s protocol implementation
Google DeepMind says researchers used a new agentic system called AlphaProof Code to formally verify Signal’s cryptographic protocol implementation, not just its design. The work, announced on November 10, 2026 by Pushmeet Kohli and Swarat Chaudhuri, is described as the first AI-assisted formal verification of a widely deployed cryptographic protocol at this depth. DeepMind says human experts translated security goals into Lean specifications, Gemini Carbon-powered agents generated more than six million lines of proofs, and the system caught two reintroduced bugs in an isolated Signal codebase.
Original sources
- Google DeepMind · Introducing AlphaProof Code: a breakthrough in verifying complex software at depth ↗
- Wired · Google DeepMind researchers used agentic AI framework AlphaProof Code to verify the Signal protocol's cryptographic guarantees, in a four-week proof-of-concept (Lily Hay Newman/Wired) ↗
- Techmeme · Google DeepMind researchers used agentic AI framework AlphaProof Code to verify the Signal protocol's cryptographic guarantees, in a four-week proof-of-concept (Lily Hay Newman/Wired) ↗