Pushmeet Kohli and Swarat Chaudhuri led a DeepMind effort that proved selected libsignal guarantees down to machine code, with humans reviewing specifications and flagged proof code.
By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff)
· Published
Primary source: [Hacker News](https://deepmind.google/blog/introducing-alphaproof-code/)
Why it matters #
Formal verification has usually been too expensive for widely deployed production code. If AI agents can produce machine-checked proofs against real cryptographic implementations in weeks, the constraint shifts from proof generation to writing precise specifications and auditing the assumptions underneath them.
Google DeepMind researchers led in the announcement by Pushmeet Kohli (@pushmeet) and Swarat Chaudhuri said on November 10th, 2026 that AlphaProof Code, a multi-agent formal verification framework, had mathematically proved selected security properties of the actual code underlying the Signal protocol. DeepMind described AlphaProof Code as combining human expertise with AI agents, and said the work marked the first use of AI to verify a widely deployed cryptographic protocol at this depth.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.