Anthropic AI unveils Claude Code Security to hunt hidden bugs

Anthropic has introduced Claude Code Security, an AI-powered service built into Claude Code that scans codebases for vulnerabilities and suggests fixes, with the capability currently rolling out in a limited research preview for Enterprise and Team customers.

According to the announcement, the tool analyses codebases like a human researcher, focusing on data flows and component interactions to uncover issues that traditional scanners might miss, and every finding undergoes multi-stage verification with a severity and confidence rating displayed on a human-review dashboard.

The release notes that Claude Code Security has been refined over more than a year of research, including testing in Capture-the-Flag events and in partnership with Pacific Northwest National Laboratory, and that Claude Opus 4.6 helped discover over 500 long-undetected bugs in open-source projects. The aim is to empower defenders to review findings and iterate on fixes directly within familiar tools, while ensuring responsible deployment. Anthropic expects AI-driven code scanning to become more widespread, but cautions that attackers will also leverage AI to find weaknesses faster.