Understand this story
On November 10, 2026, Google DeepMind announced that it had used the AlphaProof Code agent framework to complete formal verification of key cryptographic guarantees in the actual code of the Signal protocol, generating more than six million lines of Lean proofs, with the entire process taking four weeks. The framework divides work among multiple LLM agents, while human experts are responsible for determining the security goals to be proven and reviewing about one percent of the code. The team also reintroduced two real, already-fixed vulnerabilities into an isolated version of the Signal code, and the system produced machine-checked insecurity proofs for both. DeepMind also open-sourced the formal verification artifacts. On November 12, The Decoder reported that the framework is based on Gemini Carbon, that work which would normally take years was completed in four weeks, and that what was verified is the cryptographic core code actually running on users' devices; Signal CTO Ehren Kret said that amid growing AI threats, this approach strengthens trust in data privacy.
AI-generated from public reports
Latest update · Discussion
gihyo.jp joined this discussionDevelopments
1 confirmed developments
Google DeepMind uses AlphaProof Code to formally verify the Signal protocol
Source:Google DeepMind
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 13
gihyo.jpRelated discussion
gihyo.jp joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Nov 12
The DecoderSelected
Google DeepMind's AlphaProof Code completes verification of Signal's encryption core in four weeks
Google DeepMind has announced AlphaProof Code, which can mathematically prove that software meets security specifications, and has completed a real-world verification on the Signal encryption protocol. The research team said that after experts defined the formal security requirements, an AI agent based on Gemini Carbon wrote more than six million lines in the Lean proof language, with about 1% of the code flagged for human review, completing in four weeks work that would normally take years. Previous verification work mostly remained at the theoretical design stage; this time, what was verified was the encryption core code actually running on users' devices, and two bugs deliberately planted for testing were immediately found. Signal CTO Ehren Kret said that amid growing AI threats, this approach strengthens trust in data privacy; DeepMind plans to extend the method to power grids and medical devices through systems such as CodeMender.
Nov 11
36Kr · AI
DeepMind uses AlphaProof Code to generate 6.5 million lines of Lean proofs for the Signal protocol
DeepMind's AlphaProof Code generated approximately 6.5 million lines of Lean proof code for the Signal core protocol implementation within 4 weeks, covering about 140,000 machine instructions, with only about 1% of the code ultimately requiring further expert inspection. The research team also reinstalled two real bugs previously fixed by Signal into an isolated test version, and AlphaProof Code detected both and provided proofs that passed Lean checking. The relevant specifications and approximately 6.5 million lines of proof code have been open-sourced, but the verification scope is limited to a specific version of the Signal core protocol implementation and designated security properties, and does not cover the entire Signal App or all attack scenarios.
Nov 10
9to5Mac
Google DeepMind uses agentic AI to formally verify the security of the Signal protocol
Google DeepMind has completed a formal verification of the Signal protocol, mathematically proving that the protocol satisfies its security claims under all conditions. The protocol is also used by WhatsApp and Google Messages; the verification targets the protocol itself rather than the Signal app as a whole. When testing older versions, the method identified known vulnerabilities and flagged them as failing verification. DeepMind research scientist Swarat Chaudhuri said there is still a long way to go before this method can be extended to all software worldwide, and that the process is extremely compute-intensive, with it remaining unclear whether it can be used more broadly.
Unite.AISelected
Google DeepMind uses AI agents to generate Lean security proofs for Signal's compiled code
On November 10, 2026, Google DeepMind released AlphaProof Code, using AI agents to generate machine-checked Lean proofs for Signal's compiled libsignal code, covering properties such as forward secrecy, authentication, and post-compromise security. Five researchers completed the proof work within four weeks, with the agents generating approximately 6.5 million lines of Lean proof code, based on roughly 26,000 lines of trusted specifications and AArch64 semantics. The paper also notes that the proof does not cover the post-quantum recovery capability of the SPQR design, and that the computational model breaks down under large-scale messages and sessions.
Google DeepMindFirst-partySelected
Google DeepMind Formally Verifies Signal Protocol Code with AlphaProof Code
Google DeepMind researchers used the AlphaProof Code agent framework to complete formal verification of key security properties of the actual Signal protocol code, generating more than six million lines of Lean proofs, with the entire process taking four weeks. The framework divides work among multiple LLM agents, while human experts are responsible for determining the security goals to be proven and reviewing about one percent of the code. The team also reintroduced two real, already-fixed vulnerabilities into an isolated version of the Signal code, and the system produced machine-checked insecurity proofs for both. DeepMind also open-sourced the formal verification artifacts.
Wired AI
Google DeepMind Formally Verifies Signal Protocol Security with AlphaProof Code
Google DeepMind researchers used agentic AI to complete a formal verification of several key cryptographic guarantees of the Signal protocol, a project that took only four weeks and used a new AI framework called AlphaProof Code. The analysis found no new vulnerabilities in the portions examined, but on an older version of Signal containing a known old vulnerability it correctly identified the problem and showed that the protocol could not pass formal verification. The team will release the specifications and proofs used in this analysis, and DeepMind research scientist Swarat Chaudhuri said there is still a long way to go before this capability can be extended to software worldwide.
Heat trend for this story
- Current
- 9
- Period peak
- 70 Nov 12 03:00
- 24-hour change
- -58%
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.