Technology and Scienceupdated 17m ago2 min read
Google DeepMind Formally Verifies Key Signal Protocol Security With AlphaProof Code
23outlets comparedWhere the coverage agrees
13 Other3 West Asian3 Western Alternative4 Other source typesThis article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
AlphaProof Code formally verified Signal protocol's cryptographic guarantees in four weeks. Verification covered machine-code level properties for PQXDH handshake and Triple Ratchet.
Save coverage image
23 outlets in this story. Read their headlines below.
Signal Protocol, Verified
Google DeepMind used agentic AI to mathematically prove, or "formally verify," key cryptographic guarantees of the Signal protocol in a four-week proof-of-concept project. Swarat Chaudhuri said, "We are very excited by this result, but there is a long distance to go before we are able to scale this to all the software of the world, and we want to do that," describing the work as a milestone rather than a full guarantee for every part of Signal. DeepMind’s verification did not cover the entire Signal mobile app, and it did not find any new vulnerabilities in the portions of the Signal protocol it looked at.
DeepMind also used AlphaProof Code on older versions of Signal that contain known and long-fixed vulnerabilities, and the system correctly identified them and showed that the protocol could not be formally verified while those issues remained. Google DeepMind released the specification and proof it developed for the Signal protocol analysis to aid other formal verification work.
Down to Machine Code
Google DeepMind used AlphaProof Code to verify key security guarantees in the compiled code behind the Signal Protocol, including protections for encrypted conversations and messages when an attacker temporarily obtains a device’s cryptographic keys. DeepMind said its proofs covered PQXDH, the handshake that establishes a shared secret at the start of a conversation, and the Triple Ratchet, which updates the keys used to encrypt messages. DeepMind’s team checked whether those guarantees hold in the machine instructions produced when the reference implementation is compiled, where flaws can arise even when the design is sound.
Signal CTO Ehren Kret said the work expands the scope of earlier verification efforts and gives the project greater confidence in the protocol’s privacy guarantees. DeepMind’s researchers announced the results on November 10 and said the verification took four weeks, while also publishing the verification artifacts on GitHub.
What It Means Next
Google DeepMind’s AlphaProof Code process generated about 6.5 million lines of proof code in Lean, and human experts reviewed the specifications before the agents produced proofs. DeepMind said agents powered by Gemini Carbon generated about 6.5 million lines of proof code in Lean, and Lean checked whether each proof established the stated result. DeepMind’s team modeled how compiled Arm instructions run so Lean could check claims about the executable itself, and it separately proved handshake secrecy in a model where an attacker can later break the classical Diffie–Hellman exchange.
DeepMind’s researchers emphasized that their guarantees depend on attacker models and assumptions about the underlying cryptographic operations, and they said the scope was the core protocol implementation with side channels and the rest of the messaging app outside the analysis. Google DeepMind’s approach also aimed to bridge the gap between design and implementation by verifying the compiled code, and it released the artifacts publicly so other formal verification work can build on the specification and proof.