Anthropic adds security checks and company-wide controls to Claude Enterprise

Companies using Claude Enterprise can now connect their security tools to checks on prompts and tool calls. Anthropic’s November 5, 2026 security update makes those connections generally available, alongside beta controls for reviewing settings, managing company-wide access and blocking outside accounts. The package also details local activity monitoring that detects actions without stopping them.

From reviewing settings to checking actions

Anthropic describes the update as a response to security teams’ need to see how Claude is configured, what it does on company devices and how it handles data. Its inference hooks connect supported Claude products to security vendors, letting organizations apply policies they have already written rather than recreate them inside Claude.

There are two checks. Prompt hooks send prompts to the organization’s security tool. Pre-tool-call hooks check proposed tool calls before they run. Both are generally available, and connect to tools used for data loss prevention—policies designed to keep sensitive information from leaving approved boundaries.

A separate security center, now in beta, compares settings for chat, Cowork and Claude Code with Anthropic’s recommendations. Each control shows its current value, a recommended value where one exists and a link to change it. Automatically checked controls refresh when the page loads; customers with the Compliance API can retrieve the same results through it.

Admins can also record reviews, accept risks and flag issues together, then download a report for auditors or other stakeholders. The report distinguishes automatic checks from controls confirmed by an organization Owner. Anthropic cautions that the center offers guidance, not certification or a guarantee of legal or standards compliance.

One company console, with separate boundaries

Centralized management, also in beta, gives a parent organization one console for company-wide identity and access settings across its Claude Enterprise and Claude Platform organizations. It covers single sign-on, directory synchronization, domain verification, groups and Compliance API keys. Companies can connect multiple identity providers, allowing an acquired business to retain its existing provider.

Individual organizations retain their own members, roles and security controls. A parent-level role does not automatically grant access inside them, although parent-created Compliance API keys can read data across organizations. Setup requires approval from admins of the organizations involved and cannot be undone.

Other beta access controls let admins manage approved IP addresses themselves and sign a member out through one Admin API call without removing their membership. That sign-out ends browser and device sessions and requires apps such as Claude Code to authenticate again.

On managed computers, beta Claude Desktop controls let admins block personal and other outside Claude accounts. Admins can set commonly used Desktop settings without pushing a device-management update, while Owners can restrict which workspace folders people may use with Claude.

Visibility inside local task execution

The announcement also details a recently announced endpoint detection and response integration. When Cowork and unified Claude execute code locally, they use an isolated virtual machine on the user’s computer. Anthropic says security teams requested visibility inside that environment comparable to what they have elsewhere on the device.

The beta lets a supported security vendor’s sensor run inside that virtual machine and report to the organization’s existing security console. Deployment uses device-management software on macOS and Windows. Each time the virtual machine starts, Claude checks the sensor package’s vendor signature before installing it.

Those sensors run in detection mode: they report activity but do not block it. Coverage stops at the local virtual machine and does not extend to Cowork sessions in the cloud.