Skip to content
NewA calmer, faster ZeroHour
StoryAI researchinfo

Anthropic says Claude formally verified 123 open-source libraries

2 sources · 2 articlesFirst reported Updated ()1
What's new

First merged summary.

Merged summary · x-ai/grok-4.7 · rewritten as coverage arrivesSummaries by Vani AI

Anthropic and Theorem say Claude produced Rocq-checked assembly-level security proofs for 123 libraries, under narrow assumptions.

Anthropic’s Frontier Red Team and partner Theorem say Claude generated formal proofs of security properties for code from 123 open-source libraries in Google’s OSS-Fuzz project. The proofs, about 180 million lines and checked by Rocq, were produced in a few weeks at RISC-V assembly level. Anthropic says they guarantee a normal exit without memory-safety errors or failed assertions for every input under 64 MB and every modeled Linux response, assuming a single-threaded static binary. While proving, Claude found and fixed hundreds of bugs, including long-standing issues in a widely used PDF reader, according to the companies. A subset of the proofs, RISC-V and OS semantics, and a Claude skill are being published; Theorem offers the pipeline to other projects and names glibc, OpenSSL, and cURL on its roadmap.

Key facts
  • Proofs cover 123 OSS-Fuzz libraries at RISC-V assembly level.
  • About 180 million lines of proof, checked by Rocq, in a few weeks.
  • Assumptions include a single-threaded static binary and inputs under 64 MB.
  • Claude found and fixed hundreds of bugs, including a decade-old logic error.
  • A subset of proofs, semantics, and a Claude skill are being published.
  • Opus 5.5 used about 40 percent fewer tokens than Opus 5 on half the proofs.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

What each outlet adds

beyond the shared facts
  • Hacker NewsRISC-V scope, publishing a Claude skill, and Anthropic’s first large-scale claim.
  • The Decoder180 million Rocq-checked lines, OSS-Fuzz, PDF-reader bugs, Opus token cut, roadmap.
Vendors
Anthropic
Products
RocqOSS-Fuzz
Organizations
AnthropicTheoremGoogle
AI models
ClaudeOpus 5.5Opus 5

Coverage

oldest first · one row per article · UTC
  1. · 2d ago Hacker News · securityFirst report

    Anthropic says Claude proved properties in 123 open-source libraries

    Proofs cover single-threaded RISC-V static binaries and inputs under 64 MB; a subset and a Claude skill are being published.

  2. · 2d ago The Decoder

    Claude produced formal proofs for code from 123 OSS-Fuzz libraries

    Rocq checks the proofs, which hold only for one architecture, single-threaded code, and inputs under 64 MB.