Claude Security Enters Public Beta

Claude Security has moved into public beta for Claude Enterprise customers. It scans code repositories to find vulnerabilities, and can continue on to review patches proposed by Opus 4.7. It's usable directly within the Claude platform, and will also expand through partner products and service channels.

This feature is an evolution of the earlier Claude Code Security, refined through a limited research preview involving hundreds of organizations. It can be started right away without any separate API integration or building custom agents, and also supports scheduled scans and scans targeting specific directories or branches.

During scanning, Claude reads interactions between files and modules and traces data flow to find vulnerabilities. Each finding comes with confidence, severity, impact, and reproduction steps, and users can go straight on to handle the suggested fix instructions in Claude Code on the Web.

Anthropic viewed detection quality and scan-to-fix time as key metrics in particular. A multi-stage verification pipeline independently re-reviews each result to reduce false positives, and some teams completed everything from scanning to applying patches in one place.

This update also adds the following features:

• Export findings as CSV or Markdown
• Dismiss a finding while leaving a rationale
• Send results via Slack, Jira, or webhooks

CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, TrendAI, and Wiz are integrating Opus 4.7 into their security tools, and service partners such as Accenture, BCG, Deloitte, Infosys, and PwC are also helping with deployment. Opus 4.7 has also been given new safeguards that automatically block high-risk cyber requests, and support for Claude Team and Max customers is coming soon.