Google DeepMind has utilized artificial intelligence to perform a formal verification of the Signal encryption protocol. This mathematical process aims to prove the reliability of end-to-end encryption, which is also currently implemented in WhatsApp and Google Messages.
Unlike traditional testing methods that search for known vulnerabilities, formal verification uses mathematical rigor to ensure security requirements are met under all defined conditions. The AI agents successfully identified security flaws in older versions of the Signal protocol during the research.
The analysis process required approximately four weeks to complete. Google has not disclosed the specific amount of computational resources consumed during the project, noting that scaling this technology to broader software applications remains a long-term goal.
DeepMind researcher Swarat Chaudhuri stated that while the team is encouraged by the results, significant work is needed before this methodology can be applied to software on a global scale. The current study built upon existing non-AI formal verification efforts.
