Independent tech intelligence, checked against primary sources.

TechPulseMind Useful technology.
No manufactured hype.

AlphaProof Code: DeepMind AI formally verifies Signal’s crypto in four weeks

Google DeepMind's AlphaProof Code used AI agents to formally verify Signal's cryptographic code in four weeks, with Lean-checked proofs and open-sourced artifacts.

AlphaProof Code: DeepMind AI formally verifies Signal’s crypto in four weeks

Google DeepMind researchers used a new agentic framework, AlphaProof Code, to formally verify the core cryptographic code behind the Signal protocol — a task the company says took four weeks, where expert teams have historically needed years. The work, published November 10, covers the code that protects private conversations in apps including Signal and Google Messages.

According to DeepMind, it is 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, with or without AI. Formal verification means mathematically proving that code behaves as intended.

How the agents split the work

AlphaProof Code is a multi-LLM framework that divides verification the way a research team would. Human experts work with a mix of models to turn Signal’s security goals into formal mathematical specifications, with every one reviewed by a person. Agents powered by Gemini Carbon then generated over six million lines of Lean, the language used to machine-check proofs.

The agents also audited their own output and flagged areas needing a second look by a human — about one percent of the codebase, according to DeepMind. Lean automatically checks the rest.

The Register reports that the verification covered Signal’s post-quantum PQXDH handshake and Triple Ratchet, validating the design of libsignal’s forward secrecy, authentication, post-compromise security and handshake secrecy. DeepMind says its agents went beyond the design level to prove that the compiled code fulfills those properties down to the exact code running on devices.

To test whether the system can tell correct code from buggy code, the researchers reintroduced two real, previously fixed bugs into an isolated version of Signal’s code. In both cases, DeepMind says, the system produced a machine-checked proof that the buggy version was insecure.

Signal’s CTO calls it a breakthrough

Ehren Kret, chief technology officer of Signal, said the approach expands what verification can cover. “Using AI agents to translate live code into valid Lean theorems is a major breakthrough,” he said. “As AI-driven security threats grow, rigorously verifying the Signal protocol implementation gives us even greater confidence in our code’s privacy guarantees.”

Swarat Chaudhuri, a research scientist at Google DeepMind, told The Register the system is “fundamentally built for human AI collaboration” and that the next frontier is specification rather than proof. The Register notes the result is theoretical in that it assumes perfect cryptography and does not guarantee the computational secrecy of bytes.

DeepMind says it is open-sourcing its formal verification artifacts for community use, and the technical paper is available from its blog.

Related reading

Sources

Some links on this page may be affiliate links. If you buy through them we may earn a commission at no extra cost to you. See our affiliate disclosure.