Making frontier cybersecurity capabilities available to defenders
Summary
Anthropic introduces Claude Code Security, a reasoning-based static analysis tool embedded in Claude Code to detect complex vulnerabilities and propose patches for human review. Findings are verified through a multi-stage process with severity ratings and a dashboard, enforcing human approval before changes. The limited research preview targets enterprise teams and open-source maintainers to accelerate secure coding and defend against AI-enabled attacks, while acknowledging potential misuse and stressing responsible deployment.