Anthropic expands Claude Security public beta — continuous repository scanning, automatic fix tracking and editable threat models

Late October 2026 Anthropic expanded Claude Security (public beta for Claude Enterprise) with continuous repository scanning that re-scans on new commits based on risk/complexity, multi-repo and cross-service vulnerability detection, multi-stage validation, editable team-informed threat models, severity tuned to project security context, and automatic fix tracking that marks findings resolved and reopens them on recurrence. It also adds an API, webhooks and MCP integrations with role-based access control.

This turns Claude Security from a one-time scan into a live, continuously validated security layer for AI-built code — directly relevant as AI generates more application code and vulnerabilities like PHP2Shell show AI-to-exploit speedups.

Claude Enterprise customers developing with Claude Code should enable the Claude Security beta and wire its API/MCP outputs into existing code-scanning and CI workflows; non-Enterprise teams should track GA timing.