Anthropic previews Claude Code Security for AI-assisted vulnerability review
Summary
Anthropic’s limited research preview of Claude Code Security scans codebases for vulnerabilities, verifies findings, and suggests patches for human review. Developers must approve fixes before they are applied.
Key points
- Anthropic says Claude Code Security is built into Claude Code on the web and is available in a limited preview to Enterprise and Team customers; open-source maintainers can apply for expedited access.
- The tool reasons about code interactions and data flows to seek complex vulnerabilities that rule-based analysis may miss.
- Findings undergo multi-stage verification and receive severity and confidence ratings.
- Developers can inspect suggested patches, and no fix is applied without human approval.
- Anthropic says its team found more than 500 vulnerabilities in production open-source codebases using Claude Opus 4.6; triage and responsible disclosure with maintainers are ongoing.
Article Details
- Event Type
- Limited research preview launch of an AI-assisted code security vulnerability detection and patch-suggestion capability.
- Impact
- Claude Code Security scans codebases for vulnerabilities, verifies findings through multiple stages, assigns severity and confidence ratings, and suggests patches for human review. Anthropic says its team used Claude Opus 4.6 to find more than 500 vulnerabilities in production open-source codebases and is working with maintainers on triage and responsible disclosure. The article does not report exploitation or customer incidents; fixes require human approval.
Vendors
Products
Claude CodeClaude Code Security is a new capability built into Claude Code on the web and is now available in a limited research preview to scan codebases for security vulnerabilities and suggest targeted patches for human review.Claude Code SecurityClaude Code Security is a new capability built into Claude Code on the web and is now available in a limited research preview to scan codebases for security vulnerabilities and suggest targeted patches for human review.Claude Opus 4.6Using Claude Opus 4.6, released earlier this month, our team found over 500 vulnerabilities in production open-source codebases—bugs that had gone undetected for decades, despite years of expert review.