AlphaProof Code verifies Signal’s cryptographic core in four weeks
Google DeepMind has introduced AlphaProof Code, an AI system designed to mathematically verify that software conforms to defined security requirements. In a demonstration involving Signal’s encryption protocol, the system completed verification of the cryptographic core running on users’ devices in four weeks. Similar work has generally taken years.
Researchers first established formal security specifications. AI agents powered by Gemini Carbon then generated more than six million lines of proof code in the Lean language. The process identified roughly one percent of the codebase for human review. Two bugs deliberately inserted for testing were detected immediately.
According to DeepMind researchers Pushmeet Kohli and Swarat Chaudhuri, previous verification projects had not progressed beyond the theoretical design. AlphaProof Code is intended to help create specifications and then independently prove that the implementation satisfies them. Signal CTO Ehren Kret said the approach could reinforce confidence in data privacy as AI-related threats increase. DeepMind plans to apply similar methods to power grids and medical devices, including through systems such as CodeMender.
Written by Sakhanda Wire AI from reporting by The Decoder. Facts, quotes and full context are in the original article below.