Sources checked
How AlphaProof Code verified Signal
Google DeepMind researchers used a new multi-agent AI framework called AlphaProof Code to mathematically verify key cryptographic guarantees in the Signal protocol, marking the first time AI has formally verified a widely deployed encryption protocol at the machine-code level.
The project took four weeks and produced over three million lines of machine-checked proofs written in the Lean theorem prover.
AlphaProof Code is a multi-LLM agentic framework that splits formal verification work the way a human research team would. Human experts work with AI to decide what to prove, and a group of AI agents executes the proofs.
According to the DeepMind team, researchers collaborated with a mix of models to translate Signal's security goals into formal mathematical specifications, with human experts reviewing every one.
From there, agents powered by Gemini Carbon generated over six million lines of Lean code — a computer language that mathematically proves the accuracy of other code.
After generating the proofs, the agents audited the entire codebase and highlighted just one percent of it for human review. Because Lean automatically checks the rest, human experts could focus their review where judgment mattered most.
Beyond design-level verification
Most cryptographic protocol verification stops at the design level — checking that the abstract mathematical specification is sound.
AlphaProof Code went further, proving that the compiled machine code fulfills key security properties, such as keeping messages protected before and after a device is temporarily compromised.
The team verified the code running on devices, not just the design documents, bridging the gap where real-world vulnerabilities often hide.
"While prior formal verification has necessarily focused on smaller scopes, this new approach enables a significant expansion," the DeepMind team wrote.
"Traditionally, verifying that production code strictly matches its cryptographic design required extensive manual effort, making the use of AI agents to automatically translate live code into valid Lean theorems a great breakthrough."
The researchers confirmed AlphaProof Code can distinguish correct code from buggy code. They reintroduced two real, previously fixed bugs into an isolated version of Signal's codebase.
In both cases, the system produced a machine-checked proof that the buggy version was insecure.
No new vulnerabilities found
The analysis did not find any new vulnerabilities in the portions of the Signal protocol it examined.
When the researchers used AlphaProof Code on older versions of Signal containing known and long-fixed vulnerabilities, it correctly identified them and showed that the protocol could not be formally verified.
"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," Swarat Chaudhuri, a DeepMind research scientist, told WIRED.
Signal's Chief Technology Officer Ehren Kret said in a statement: "While prior formal verification focused on smaller portions of the code, Google DeepMind's approach enables significant expansion.
As AI-driven security threats grow, rigorously verifying the Signal protocol implementation gives us even greater confidence in our code's privacy guarantees."
What was verified and what was not
DeepMind's analysis focused on verifying the core cryptographic protocol implementation — the mathematical guarantees that underpin Signal's encryption.
The team did not attempt to formally verify the entire Signal mobile application, the operating system it runs on, or non-cryptographic application logic. The researchers are releasing their formal specification and proof artifacts for community use.
Pushmeet Kohli, chief scientist at Google Cloud and DeepMind vice president of research, told WIRED that the breakthrough opens the door for banks and other enterprises that "rely on assurances and who worry about bugs" to mathematically prove the security of their systems.
Cost and scalability questions remain
DeepMind conducted the Signal analysis as a research project and declined to share its cost or whether it plans to release a commercial formal verification product.
The open question across the industry is whether AI-automated formal verification can scale at a workable cost.
Chaudhuri emphasized the importance of keeping humans in the loop. "We don't want this sort of vibe world where AI is generating the goals and AI is generating the proofs and basically taking tests that are self-written," he said.
"There's a middle ground."
Wider implications for software security
The team described the approach as a foundation that could eventually be applied to other critical systems beyond messaging encryption.
"Verifying Signal's core cryptography lays the groundwork to secure other vital systems, from medical device firmware to power grid software," the researchers wrote.
The project builds on earlier formal verification work by the Signal Foundation and on DeepMind's AlphaProof research, but represents a step change in scale — from isolated proofs to millions of lines of machine-checked code covering an entire production protocol.
The researchers are open-sourcing their formal verification artifacts to accelerate research in the field.
Article sources
* Google DeepMind — Introducing AlphaProof Code: Formal Verification of Signal's Protocol with Gemini (Pushmeet Kohli, Swarat Chaudhuri, November 10, 2026) * WIRED — A New AI Tool From Google DeepMind Proved the Signal Protocol Is Secure (Lily Hay Newman, November 10, 2026)
Sources & further reading
How this story was made
Written by Mustak4 with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. The cover is an AI-generated editorial illustration, not a photograph or evidence of an actual event.
Our editorial standards
