Introducing AlphaProof Code: a breakthrough in verifying complex software at depth

2026-11-13 · Google DeepMind

Introducing AlphaProof Code: A Breakthrough in Verifying Complex Software at Depth

Google DeepMind has announced a milestone in software security long considered a grand challenge: researchers successfully guided AI agents to mathematically prove key security properties of the actual code underlying the Signal protocol, the technology protecting billions of private conversations in apps like Signal and Google Messages.

This breakthrough was achieved using AlphaProof Code, a novel agentic framework for formal verification at unprecedented depth. Formal verification—mathematically proving code runs correctly without flaws—is the highest standard of security assurance, though it is traditionally difficult and labor-intensive. This marks the first time AI has been used to formally verify a widely deployed cryptographic protocol at this depth, and the first time such a protocol's implementation has been formally verified by any method.

How Signal Was Verified

AlphaProof Code is a multi-LLM agentic framework that divides formal verification tasks similarly to a research team: human experts collaborate with AI to decide what to prove, and AI agents execute the proofs.

  • Specification Translation: Researchers collaborated with various models to translate Signal’s security goals into formal mathematical specifications, with human experts reviewing each one.
  • Proof Generation: Agents powered by Gemini Carbon executed the proofs, generating over six million lines of Lean, a computer language that mathematically proves code accuracy.
  • Auditing and Review: The agents audited the proof and highlighted areas needing human review—just 1% of the codebase, saving massive time. Lean automatically checked the rest, allowing experts to focus their judgment where it mattered most.

Significance and Complexity

Cryptographic protocols underpin nearly all digital communication, and a single flaw can expose millions. While experts have spent decades formally analyzing them, most verification stops at the design level.

AlphaProof Code's agents went further, proving that the compiled code fulfills key security properties down to the exact code running on devices. Because real-world vulnerabilities often hide in the gap between design and implementation, bridging this gap makes proofs vastly more complex. Despite this complexity, the verification process took just four weeks—a task that historically takes expert teams years.

Testing the System

To confirm AlphaProof Code can distinguish correct code from buggy code, researchers reintroduced two real, previously fixed bugs into an isolated version of Signal's code. In both cases, the system produced a machine-checked proof demonstrating that the buggy version was insecure.

Industry Feedback and Future Outlook

Ehren Kret, Chief Technology Officer of Signal, noted that while prior formal verification focused on smaller scopes, this AI-driven approach enables significant expansion. Automatically translating live code into valid Lean theorems provides greater confidence in the code's privacy guarantees as AI-driven security threats grow.

Achieving this challenge demonstrates how AI can bring elite security assurance to businesses, governments, and critical infrastructure. Verifying Signal's core cryptography lays the groundwork to secure other vital systems, from medical device firmware to power grid software.

Scaling these techniques requires continued research. Paired with automated defense agents like CodeMender, formal verification can counter modern cyber threats. Google DeepMind has open-sourced its formal verification artifacts for community use.

Source