On 10 November 2026 Google DeepMind announced AlphaProof Code, an agentic framework for formal verification: mathematically proving that code behaves as specified. Its first showcase is the implementation of Signal's cryptographic protocol, the encryption used by the Signal messenger.
How it worked, per DeepMind: human experts work alongside several AI agents. Agents powered by Gemini Carbon generated more than six million lines of Lean proof code; they flagged about 1% of it for human review, and the Lean proof checker verified the rest. The job took four weeks, work that DeepMind says has historically taken expert teams years. Runtime Wire reports the effort was led by Pushmeet Kohli and Swarat Chaudhuri.
Testing the tester: the team put back two bugs that Signal had fixed in the past. For the buggy versions, AlphaProof Code produced machine-checked proofs that they were insecure, showing the method catches real flaws rather than proving anything it is given.
What it claims, and what it doesn't: DeepMind calls this the first time AI has formally verified a widely deployed cryptographic protocol at this depth, and the first time such a protocol's implementation has been formally verified by any method. Runtime Wire notes that only selected guarantees were proven, and that the hard part shifts to writing precise specifications and checking the assumptions under them. DeepMind is open-sourcing the verification artifacts and a technical paper; it has not said whether AlphaProof Code itself will be released or who can use it.
