Skip to content
Live
Trending storyPast story

Claude completes formal verification of 239 entry point assembly layer

1 reports1 outlets28 days agoUpdate

The story so far

AI overview

Anthropic partnered with Theorem to use Claude to complete formal proofs for the RISC-V assembly layer of 123 open-source libraries with 239 entry points from OSS-Fuzz, proving that these programs will not crash with any input less than 64 MB, with the proof code being approximately 180 million lines of Rocq. Anthropic claims this indicates that LLMs are now capable of producing formal proofs at scale. The conclusion comes with premises: the program must run as a single-threaded static binary, with a fixed memory budget, on a RISC-V 1 CPU and a Linux system; the proof covers all inputs less than 64 MB under the above conditions and the modeled operating system responses and memory layout. Anthropic has made some of the proofs and the RISC-V and operating system semantics available for others to verify and reuse, but has not disclosed the proofs that may expose unpatched vulnerabilities. Theorem states that it can run the same process for other open-source projects and has released a Claude skill for developers to verify their own code.

AI-written from the reports · Updated 26 days ago

Coverage timeline

Follow the coverage to see every side of the story.

22 Oct
  1. Anthropic ResearchFeatured
    LLMs are making formal software verification feasible at a much larger scale

    Anthropic partnered with Theorem to complete formal verification of the RISC-V assembly layer for 239 entry points in 123 OSS-Fuzz open-source libraries using Claude. The verification proves that these programs will not crash with any input less than 64 MB, with the proof code consisting of about 180 million lines of Rocq.