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

Anthropic's Claude Opus 4.6 identified and helped validate over 500 previously unknown high-severity security flaws in open-source projects (including Ghostscript, OpenSC, and CGIF); maintainers have since patched the reported defects. The report highlights the model's ability to reason about code and find complex memory- and logic-related bugs, and warns that while this aids defenders, similar AI capabilities could lower barriers for misuse in multi-stage cyber attacks.