Anthropic used Claude to complete assembly-level formal security proofs for 123 open-source libraries
LLMs are making formal software verification feasible at a much larger scale
AI BriefingAnthropic partners with Theorem to use Claude to turn OSS-Fuzz fuzz tests into specifications, proving that 239 entry points across 123 open-source libraries do not crash on any input under 64 MB and generating approximately 180 million lines of Rocq proofs.
Also reported by 2 other sources
- Anthropic uses Claude to automatically generate mathematical proofs to verify the security of existing codeThe Decoder:AI News · 24 days ago
- Anthropic published a blog post on using Claude to formally verify open-source software security at scaleAnthropic · 25 days ago
Source: Anthropic: Research · anthropic.com
Event follow-up1
View all event coverage- Anthropic used Claude to complete assembly-level formal security proofs for 123 open-source librariesAnthropic:Research · 25 days ago