This diagram shows how the AI’s work gets checked.

Google DeepMind says AI verified key security properties of Signal’s machine code

Google DeepMind says AlphaProof Code used models including Gemini Carbon to formally verify code running on devices for the Signal protocol. Humans set the security goals and Lean checked the AI-generated proofs. When researchers reintroduced two previously fixed bugs, the system produced machine-checked proofs that the code was insecure both times. The team is open-sourcing its specifications and verification artifacts.

1 more source reported this

• Rohan Paul (@rohanpaul_ai)

Same event; featured as “Google DeepMind 用 AI 验证 Signal 协议生产代码的关键安全属性”