跳到正文
原文
The Decoder:AI News· Manuel Uth·· 5 天前AI 评分56

Google DeepMind 发布 AlphaProof Code,四周验证 Signal 加密核心代码

Google Deepmind's AlphaProof Code verified Signal's cryptographic core in four weeks instead of years

文章要点
  • 1Google DeepMind 发布 AlphaProof Code,可数学证明软件代码符合安全规格
  • 2在 Signal 加密协议上,系统用四周完成了通常需数年的验证,且覆盖用户设备上实际运行的核心代码
  • 3基于 Gemini Carbon 的 AI 智能体写了超 600 万行 Lean 证明语言,约 1% 代码库被标记供人工审查
  • 4测试中故意植入的两个漏洞被立即发现;DeepMind 计划通过 CodeMender 把方法扩展到电网与医疗设备
决策要点

AI 把形式化验证从理论设计阶段推进到实际运行代码,将按年计的安全审计周期压缩到周级。

  • 影响:高保证软件的安全审计周期与人力成本有望显著压缩,涉及加密通信、支付类系统的企业可关注;当前公开案例限于 Signal 加密核心,向电网与医疗设备的扩展仍在计划阶段。
  • 参考:可考虑盘点自家依赖加密或安全关键代码的系统清单,评估此类验证工具成熟后的试点场景;也值得先测算引入形式化验证所需的人才与流程成本。
  • 依据:四周耗时、600 万行 Lean 代码与漏洞捕获均出自 DeepMind 研究者博客及 Signal CTO 表态,属官方口径,暂无独立验证。
正文

Manuel Uth

Nov 11, 2026

Google Deepmind has unveiled AlphaProof Code, an AI system that mathematically proves software meets its security specs without errors. Researchers demonstrated the technology on Signal's encryption protocol, completing in four weeks work that typically takes years. Earlier verification efforts stopped at the theoretical design stage, according to a blog post by DeepMind researchers Pushmeet Kohli and Swarat Chaudhuri. AlphaProof Code verified the actual cryptographic core code running on users' devices. After experts defined formal security requirements, AI agents built on Gemini Carbon wrote over six million lines in the Lean proof language, flagging roughly one percent of the codebase for human review. Two bugs planted deliberately for testing were caught immediately.

AI agents in AlphaProof Code help create the specifications and then independently prove that the program code meets them. | Image: Google Deepmind

Signal CTO Ehren Kret says the approach strengthens trust in data privacy as AI threats grow. DeepMind plans to extend the method to power grids and medical devices through systems like CodeMender.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

来源:The Decoder:AI News · the-decoder.com