Anthropic launches Claude Code Security for autonomous vulnerability scanning

Anthropic has just announced Claude Code Security, an autonomous vulnerability detection tool built into Claude Code. It scans codebases for security issues and proposes targeted fixes for human review. The rollout follows reports that attackers, including state sponsored actors, have used Anthropic models to identify exploitable vulnerabilities in both new and well maintained codebases.

Anthropic says the tool differs from conventional security products by reasoning through code more like a security researcher, rather than relying on rule based pattern matching against known vulnerability signatures. Findings include severity ratings for prioritization and confidence scores, with automated rechecks intended to cut down false positives before issues appear in a dashboard for security teams.

Claude Code Security does not change code directly and is currently offered in a limited research preview for Enterprise and Team customers, with free expedited access available to eligible open source maintainers. The announcement also appeared to pressure cybersecurity stocks, with media outlets like SiliconANGLE reporting that CrowdStrike closed down almost 8% and Cloudflare fell just over 8% in the same trading session.