{"rewrite":{"id":"r_f790abc8e9932f72159db0fd","clusterId":"c_c7abd99f0f056546c213f1a1","slug":"deepmind-uses-alphaproof-code-to-verify-the-signal-protocol-s-security","model":"deepseek-v4-1-flash","headline":"DeepMind Uses AlphaProof Code to Verify the Signal Protocol's Security","summary":"Google DeepMind's research team used an AI framework called AlphaProof Code to formally prove that the Signal protocol, used for encrypted communications in Google Messages and elsewhere, has sufficient security protections. Agents running Gemini 4 Carbon generated 6 million lines of machine-checked Lean proofs. About 1% of the codebase needed human review. DeepMind says the work, normally years of expert manual verification, took four weeks.","whyItMatters":"The team reinserted two already-fixed bugs and AlphaProof Code flagged the buggy version as not secure, which is the concrete evidence that this kind of AI proof work can catch real defects rather than only confirm code that already passed review.","webCardHtml":"\u003cp\u003eThe proof effort went down to compiled device code, the layer where DeepMind says real bugs hide. Multiple AI models first translated the protocol\u0026#39;s security goals into formal mathematical specifications, and human experts reviewed those specifications before any proof work began. Agents then carried out the proofs in Lean, a language for mathematically proving code correctness, and flagged the spots that needed human visual inspection, which came to roughly 1% of the total. DeepMind open sourced the results for the security community.\u003c/p\u003e","blueskyPost":"AlphaProof Code left about 1% of the Lean proofs for human review. DeepMind's four-week run moves the bottleneck from writing proofs to checking the small fraction the agents could not settle.","twitterPost":"AlphaProof Code cut Signal verification to four weeks, but roughly 1% of the Lean proofs still needed human review. That fraction is the new bottleneck.","threadsPost":"Google DeepMind's AlphaProof Code generated 6 million lines of machine-checked Lean proofs for the Signal protocol in four weeks. About 1% needed human review. The scarce skill shifts from writing proofs to auditing the small fraction the agents could not close.","newsletterBlurb":"Google DeepMind says its AlphaProof Code framework formally verified the Signal protocol down to compiled device code, with Gemini 4 Carbon agents generating 6 million lines of Lean proofs. Roughly 1% of the codebase needed human review. To test the tool, the team reinserted two already-fixed bugs and it flagged the buggy version as not secure. The results are open sourced.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261111-google-deepmind-alphaproof-code-verification/\",\"title\":\"Google DeepMind Successfully Formally Verifies the Security of the Signal Protocol Using Its AI Code-Verification Tool \\\"AlphaProof Code\\\"\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":5527,"outputTokens":686,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1794383363,"createdAt":"2026-11-11T07:43:46.000Z","publishedAt":"2026-11-11T07:47:11.000Z","updatedAt":"2026-11-11T07:47:11.000Z"},"cluster":{"id":"c_c7abd99f0f056546c213f1a1","canonicalTitle":"Google DeepMindがAIでコードを検証する「AlphaProof Code」を使ってSignalプロトコルの安全性の形式的検証に成功","representativeArticleId":"a_fe1f00cef789d0f5e49a705e","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[\"Google DeepMind\"],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-11-11T07:05:00.000Z","lastSeenAt":"2026-11-11T07:05:00.000Z","updatedAt":"2026-11-11T07:47:11.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20261111-google-deepmind-alphaproof-code-verification/","title":"Google DeepMindがAIでコードを検証する「AlphaProof Code」を使ってSignalプロトコルの安全性の形式的検証に成功"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":["Google DeepMind"],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":["Google DeepMind used its AlphaProof Code framework to formally verify the Signal protocol down to compiled device code.","Agents running Gemini 4 Carbon generated 6 million lines of Lean proofs.","About 1% of the codebase required human review, with Lean checking the rest.","DeepMind says the verification took four weeks rather than years of expert manual work.","The team reinserted two already-fixed bugs, and AlphaProof Code reported the buggy version as not secure."],"pageLinks":{"healthSlugs":["google-deepmind"],"storySlugs":[],"entities":[{"slug":"google-deepmind","supported":4,"disputed":0,"contradicted":0,"unverified":13,"overall":"green","story_count":11,"has_dossier":true}],"stories":[],"entitiesOk":true,"storiesOk":true}}