CLAUDE Opus 4.6, the latest large language model from Anthropic, has found more than 500 previously unknown high-severity security flaws in open-source libraries, including Ghostscript, OpenSC and CGIF. According to Anthropic, the model is not only able to identify issues but also to prioritise the most severe memory corruption vulnerabilities identified.
The tool was launched on Thursday and, prior to its debut, Anthropic’s Frontier Red Team tested it in a virtualised environment with tools such as debuggers and fuzzers to gauge its out-of-the-box capabilities. Some of the flagged defects have since been patched by their maintainers, with examples including a crash in Ghostscript from a missing bounds check and a heap buffer overflow in CGIF.
Anthropic said Opus 4.6 reads and reasons about code like a human researcher, spotting patterns and inputs likely to trigger bugs without task-specific prompting.