DeepMind AlphaProof Code verified Signal's crypto core in four weeks
Original title: Google Deepmind's AlphaProof Code verified Signal's cryptographic core in four weeks instead of years
Gemini Carbon agents wrote over six million Lean lines and sent about one percent of the code to human review.
Google DeepMind unveiled AlphaProof Code, which builds formal specifications and proves that program code meets them. On Signal, it checked the cryptographic core that runs on users' devices, work DeepMind says usually takes years. After experts set security requirements, the system flagged roughly one percent of the codebase; two bugs planted for testing were caught immediately. DeepMind says it plans to extend the approach to power grids and medical devices through systems such as CodeMender.
Unconfirmed: The claim that the work normally takes years comes from DeepMind's blog, not an independent audit.
- Earlier verification efforts stopped at the theoretical design, DeepMind researchers wrote.
- About one percent of the verified codebase was flagged for human review.
- Two bugs deliberately planted for testing were caught immediately.
- DeepMind plans later use on power grids and medical devices via CodeMender.
- Products
- AlphaProof CodeSignalCodeMender
- Organizations
- Google DeepMind
- AI models
- Gemini Carbon
From the original article
1 min read at the-decoder.com
Google Deepmind's AlphaProof Code mathematically verified Signal's cryptographic core code in four weeks, a process that typically takes years. AI agents powered by Gemini Carbon wrote more than six million lines of proof code and flagged only about one percent for human review. Deepmind plans to extend the method to critical infrastructure like power grids and medical devices.
By Manuel Uth · the-decoder.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.