Anthropic Adds New Security Controls for Claude Enterprise Administrators

Anthropic added new controls to Claude Enterprise on Thursday, Nov. 5, giving company administrators ways to review security settings, check prompts and tool requests, and coordinate policies across organizations.

The security center brings settings for Claude chat, Cowork and Claude Code into one view. It checks those configurations against Anthropic’s recommended setup, flags differences and links administrators to relevant controls. Administrators can download a report that separates automatic checks from those confirmed manually. The page refreshes when loaded and works with the Compliance API for external reporting; it offers guidance but does not guarantee compliance with legal or regulatory standards.

Inference hooks, now generally available, connect Claude with an organization’s existing data security tools. Prompt hooks send a user’s prompt to a security tool for review, while pre-tool-call hooks check a requested tool action before Claude runs it. The connections use signed HTTPS POST requests and can extend data loss prevention policies to Claude.

A parent-level console lets administrators coordinate company-wide settings across multiple Claude Enterprise and Claude Platform organizations. Its features include single sign-on, SCIM directory synchronization, domain verification, group management and oversight of Compliance API keys. Companies can connect multiple identity providers, allowing acquired businesses to keep their existing systems or transition over time. Parent-level roles provide visibility into overarching settings but do not grant access to individual organizations by default.

Other controls let administrators sign out members through the Admin API, maintain IP allowlists, block personal or external Claude accounts on managed devices, and restrict access to workspace folders.

Anthropic also added an endpoint detection and response (EDR) integration for Cowork and unified Claude. On macOS and Windows, organizations can place vendor-signed EDR sensors in an isolated virtual machine and send activity to their existing security console. The sensors detect and report activity but do not block actions. Cloud-based Cowork sessions are not covered.