---
id: "2026-11-10-deepmind-alphaproof-code-signal-verification"
url: "https://postcutoff.com/e/2026-11-10-deepmind-alphaproof-code-signal-verification/"
as_of: "2026-11-13T13:49:00+01:00"
date: "2026-11-10"
date_precision: day
category: research
importance: 4
confidence: high
status: [Confirmed]
sources: 5
editor: human editor
human_review: null
version: null
---

As of: 2026-11-13 13:49 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/e/2026-11-10-deepmind-alphaproof-code-signal-verification/

# Google DeepMind's AlphaProof Code formally verifies Signal's protocol implementation in Lean

On 10 November 2026 Google DeepMind introduced AlphaProof Code, a multi-agent framework built on Gemini Carbon that wrote over six million lines of Lean to prove that Signal's compiled protocol code, not only its design, meets key security properties. DeepMind calls it the first formal verification of a widely deployed cryptographic protocol's implementation by any method; it took about four weeks, with humans writing and reviewing the specifications.

## Key facts
- Hacker News (10 Nov): 'AlphaProof Code' reached 674 points and 197 comments; the most-discussed comment doubts the zero-bug finding ('They took a non-trivial piece of human-written code that has had bugs in the past, formally verified the whole library, and found zero bugs?'), others ask whether the specification itself is checked
- Agents powered by Gemini Carbon wrote the proofs; human experts translated Signal's security goals into formal specifications and reviewed each one; about 1% of the codebase was flagged for human review, Lean checked the rest
- Over six million lines of Lean; about four weeks, which DeepMind compares with years for expert teams on comparable work
- Properties include message protection before and after a temporary device compromise; two previously fixed bugs, reintroduced as a test, each produced a machine-checked proof that the buggy version was insecure
- Verification artifacts open-sourced at google-deepmind.github.io/code-verification; DeepMind does not say whether the framework itself is released
- Signal CTO Ehren Kret: 'Using AI agents to translate live code into valid Lean theorems is a major breakthrough.'
- Wired describes it as a four-week proof of concept

## What happened

DeepMind extended its AlphaProof line from mathematics to software: a framework of several LLM agents that turns real code into Lean
theorems and proves them. The showcase is Signal, whose end-to-end encryption protects messages for WhatsApp and Signal users. Humans
still set what "secure" means (the specifications); the agents did the proving.

## Why it matters

Formal verification of production code has been limited by the cost of writing proofs. A four-week, machine-checked proof of a deployed
protocol's implementation is a concrete case of AI agents making high-assurance software cheaper, the defensive side of the same
capabilities that worry cyber-security experts. Claims are DeepMind's; the open artifacts allow outside checking.

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 194 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 133 days after its cutoff
- Gemini 4 Hydrogen (training cutoff March 2026): 224 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 163 days after its cutoff

## Sources

1. [Google DeepMind: Introducing AlphaProof Code](https://deepmind.google/blog/introducing-alphaproof-code/) (deepmind.google, official)
2. [Wired: A new AI tool from Google DeepMind formally verifies Signal](https://www.wired.com/story/a-new-ai-tool-from-google-deepmind-formally-v) (wired.com, press)
3. [Ars Technica: Google DeepMind just mathematically verified key parts of Signal](https://arstechnica.com/ai/2026/11/signals-encryption-just-passed-an-ai-powered-mathematical-checkup/) (arstechnica.com, press)
4. [Hacker News discussion: AlphaProof Code (674 points)](https://news.ycombinator.com/item?id=50439193) (news.ycombinator.com, discussion)
5. [Google DeepMind on X: Signal verified down to the code level (140k views)](https://x.com/GoogleDeepMind/status/2120484246401777737) (x.com, discussion)

## Changes

- 2026-11-13: DeepMind X post and Ars Technica (news agent)
- 2026-11-12: Added the Hacker News thread (674 points, 197 comments) and its main sceptical question (science agent)
- 2026-11-10 (filed): Created

## Related

- 2026-10-28: [Google announces Gemini 4 Carbon, #1 on the Artificial Analysis Intelligence Index v5 (59), two weeks after Argon; first only for Fairwind cyber defenders](https://postcutoff.com/e/2026-10-28-gemini-4-carbon/index.md)
- 2026-10-08: [Google DeepMind's AlphaProof Nexus paper published in Science](https://postcutoff.com/e/2026-10-08-alphaproof-nexus-science-paper/index.md)