Google DeepMind's AlphaProof Code, Powered by Gemini Carbon, Formally Verifies Signal's Encryption in Four Weeks

On November 10, 2026, Google DeepMind introduced AlphaProof Code, an agentic framework powered by Gemini Carbon, which formally verified the cryptographic core of the Signal protocol in just four weeks. This marks the first time a widely deployed cryptographic protocol's implementation has been formally verified at this depth, with AI agents generating over six million lines of Lean proofs. For broader context, explore our Top 100 AI Tools.

AlphaProof Code: A New Approach to Formal Verification

AlphaProof Code represents a significant advancement in the field of formal verification. Unlike previous efforts that often focused on protocol design, this framework verified the compiled cryptographic code as it runs on users' devices. This distinction is crucial because it addresses the actual implementation rather than just the theoretical design, providing a higher level of assurance regarding the software's security.

The system's core relies on AI agents, specifically those powered by Gemini Carbon, to generate the extensive proofs required for verification. These proofs, written in the Lean language, were then automatically checked by the Lean prover. This automated checking process was highly efficient, flagging only about 1% of the codebase for human review, demonstrating the system's capability to handle complex verification tasks with minimal human intervention.

Accelerating Security Assurance for Signal

The verification of Signal's cryptographic core by AlphaProof Code was completed in a remarkably short timeframe of four weeks. DeepMind noted that similar tasks have historically taken expert teams years to accomplish. This acceleration in the verification process could have broad implications for the development and deployment of secure software, potentially reducing the time and resources required to achieve high levels of security assurance.

To validate its effectiveness, two real, previously fixed bugs were intentionally reintroduced into the Signal codebase during the verification process. AlphaProof Code successfully produced machine-checked proofs demonstrating that these buggy versions were insecure, confirming its ability to identify vulnerabilities. Human experts played a vital role in defining and reviewing the formal security specifications, with input from Signal's Rolfe Schmidt, Michael Rosenberg, and CTO Ehren Kret.

Implications for Cryptographic Protocol Security

The successful formal verification of Signal's encryption by AlphaProof Code sets a new precedent for ensuring the security of widely used cryptographic protocols. By verifying the actual compiled code, the system offers a robust method for identifying potential vulnerabilities at a deeper level than previously possible. This development could lead to increased trust in secure communication platforms and influence future practices in software development and security auditing.

DeepMind has also open-sourced its formal verification artifacts, which could foster further research and adoption of similar AI-powered verification techniques across the industry. This move aligns with a broader trend towards leveraging AI news and advanced computational methods to enhance cybersecurity measures.

Conclusion

Google DeepMind's AlphaProof Code, powered by Gemini Carbon, has formally verified the cryptographic core of the Signal protocol in approximately four weeks. This achievement, announced on November 10, 2026, marks a significant milestone as the first time a widely deployed cryptographic protocol's implementation has undergone such a deep formal verification. The ability of AI agents to generate millions of lines of proofs and the Lean prover to automatically check them demonstrates a powerful new approach to ensuring software security, potentially transforming how complex systems are validated for cryptographic integrity.

Sources

• Formal Verification of Signal’s Protocol with Gemini — Google DeepMind
• Verifying Signal's Security at the Machine-Code Level
• code-verification/docs/PAPER_MAP.md at main
• Gemini 4 Carbon is now available
• Google DeepMind just mathematically verified key parts of Signal