AlphaProof Code

Google DeepMind researchers have achieved a groundbreaking milestone in software security by using AI agents to mathematically prove key security properties of the Signal protocol, a technology that safeguards private conversations in apps like Signal and Google Messages. This is the first time AI has been employed to formally verify a widely deployed cryptographic protocol at such a depth, and the first instance of any method verifying the implementation of such a protocol.

The researchers utilized AlphaProof Code, a novel agentic framework for formal verification that combines human expertise with AI agents. By splitting the work of formal verification, human experts and AI agents collaborated to translate Signal's security goals into mathematical specifications, generate proofs using the computer language Lean, and audit the proofs.

This breakthrough reduces the time required for this complex task, which historically would take years to complete, to just four weeks. The verification process confirms that the core cryptographic code of Signal delivers the specified security protections. This achievement highlights how AI can provide elite security assurance to businesses, governments, and critical infrastructure, helping defend against increasingly sophisticated cyber threats.