Claude Security, a new cybersecurity tool from Anthropic, has been launched in public beta, specifically aimed at Claude Enterprise customers. The tool leverages the advanced Claude Opus 4.7 AI model to identify and address software vulnerabilities. Users can start utilizing this tool immediately without needing custom agent builds or API integrations.
Distinct from conventional methods, Claude Security mimics human-like analysis by examining the interactions of code components across files and modules. It traces data flows and interprets source code to deliver thorough insights on vulnerabilities, including their severity and potential impacts. The system also generates targeted patch instructions and assigns confidence ratings to assist in prioritizing issues.
The rollout for Claude Team and Max customers is anticipated soon, with access available through the Claude.ai sidebar or at claude.ai/security. Feedback from a limited preview involving hundreds of organizations led to various enhancements, including a multi-stage validation process that minimizes false positives. Additional functionality allows for scheduled scans and integration with tools like Slack and Jira, enabling efficient documentation and review of findings.