Anthropic launches Claude Security public beta for AI-powered code vulnerability detection

Anthropic has launched Claude Security in public beta for Claude Enterprise customers, bringing AI-powered vulnerability detection directly into production codebases without requiring custom tooling or API integrations. Powered by Opus 4.7, the platform scans code, validates findings to reduce false positives, and generates suggested patches for developers to review before deployment.

Based on feedback from its research preview earlier this year, Anthropic has added scheduled scans, directory-level targeting, CSV and Markdown exports, webhook notifications, and persistent dismissals that carry ignored findings across future scans. These updates are meant to reduce alert fatigue and make the tool easier to fit into existing security reporting and monitoring processes.

The launch follows Anthropic’s more restricted rollout of Mythos Preview on April 7, a model the company described as too dangerous for a public release and limited to roughly 50 companies through Project Glasswing. While Mythos remains tied to national security concerns and ongoing access discussions, Claude Security is positioned as a lower-friction enterprise tool for expanding vulnerability coverage without adding headcount or building internal AI infrastructure.