Cyber Risks · The Hacker News

Claude Opus 4.6 Finds 500+ High-Severity Flaws Across Major Open-Source Libraries

Summary: Anthropic said its Claude Opus 4.6 model found more than 500 previously unknown high-severity flaws in open-source libraries including Ghostscript, OpenSC, and CGIF, and that it validated each finding to rule out hallucinations. The model was tested by Anthropic's Frontier Red Team in a virtualized environment with debuggers and fuzzers but no instructions on their use, and the flagged defects have since been patched by maintainers.

What's compelling: Anthropic notes that the CGIF bug required a specific sequence of operations that could evade even a fuzzer with 100% line- and branch-coverage, so coverage metrics alone will not tell you whether a model like this has already read your dependencies' code.

Read the original at The Hacker News ↗

Topics
AI Security
Related