Google DeepMind’s AlphaProof Code formally verifies Signal’s cryptographic protocol implementation
Researchers used a new AI agentic framework to mathematically prove security properties of Signal’s code, a first for a widely deployed protocol.
1 source · cross-referenced
- AlphaProof Code is a multi-agent framework combining human expertise with AI to formally verify software correctness.
Google DeepMind researchers introduced AlphaProof Code, a multi-agent framework designed to formally verify the correctness of software implementations at unprecedented depth. The system combines human expertise with AI agents to translate security goals into formal mathematical specifications and then generate proofs of correctness.
In a collaboration with Signal, researchers used AlphaProof Code to formally verify key security properties of Signal’s cryptographic protocol implementation. The agents generated over six million lines of Lean code to prove properties such as message protection before and after device compromise, covering the exact code running on devices.
The verification process took four weeks, a task typically requiring expert teams years to complete. The framework autonomously audited the proofs, flagging only 1% of the codebase for human review, with the rest verified automatically by Lean.
To validate the system’s ability to detect flaws, researchers reintroduced two previously fixed bugs into an isolated version of Signal’s code. AlphaProof Code produced machine-checked proofs that the buggy versions were insecure, demonstrating its utility in identifying real vulnerabilities.
Ehren Kret, Signal’s Chief Technology Officer, stated that the approach enables verification at a scale previously infeasible and provides greater confidence in the protocol’s privacy guarantees amid growing AI-driven security threats.
- Nov 11, 2026 · The Verge — AI
Google DeepMind unveils AlphaGenome 2 to predict genetic variant effects on cell types
Trust79 - Nov 11, 2026 · Google DeepMind — Blog
Google DeepMind unveils AlphaGenome 2 to predict DNA variant effects in specific cell types
Trust84 - Nov 11, 2026 · arXiv cs.AI
Researchers demonstrate auditable human-agent teams for autonomous synchrotron X-ray experiments
Trust79