Google DeepMind formally verifies Signal protocol with AI

The DeepMind project builds on past formal verification work by the Signal Foundation, which the researchers say gave them a head start. While prior efforts focused on smaller portions of code, the new approach using AlphaProof Code enabled a significant expansion, according to Signal CTO Ehren Kret. The work is part of a broader push toward AI-aided formal verification, which proponents have long called a holy grail of software security because it uses math to check that software functions as intended. However, the analysis was limited in scope—it did not formally verify the entire Signal mobile app—and DeepMind researchers emphasize that scaling to all software will take time. A key open question is whether AI-automated formal verification can be practical and cost-effective at scale; Google declined to share the cost of the Signal analysis or comment on commercialization plans. The researchers also highlighted a human-AI collaboration process that keeps humans in the loop at key junctures, which Chaudhuri described as a middle ground between full AI autonomy and traditional methods.