Claude Opus 4.6 Detects 500+ Security Flaws

Anthropic revealed that its newly launched large language model, Claude Opus 4.6, detected more than 500 previously unknown high-severity security flaws across open-source libraries, including Ghostscript, OpenSC, and CGIF. Launched on Thursday, Opus 4.6 includes improved coding abilities such as code review and debugging that Anthropic says enabled automated vulnerability discovery. The findings affect widely used components and may require prompt maintainer remediation.