Anthropic makes Claude Enterprise inference hooks generally available, adds Security Center and EDR integration in beta

Anthropic announced on November 5, 2026 an expansion of security and administration features for Claude Enterprise. Inference hooks, which connect Claude to existing DLP tools, are now generally available. Most other additions are in beta. The details were reported in Japanese by KeyMan's Net (ITmedia).

What shipped

Inference hooks (generally available). Hooks let companies connect Claude to data security tools they already run, applying DLP and security policies to prompts and tool calls. There are two types: a prompt hook, which sends input prompts to a security tool for inspection, and a pre-tool-call hook, which inspects a call before the tool runs. Administrators can apply policies they have already written without recreating them in Claude.

Security Center (beta). A new page in organization settings lists security settings for chat, Cowork and Claude Code in one view. For each control it shows the current value, Anthropic's recommended value and a link to the change screen. It also surfaces new or recently changed settings. Settings are checked automatically on every page load, and organizations that use the Compliance API can retrieve the same results through the API.

An attestation feature gives administrators a shared workspace for security reviews, where they can record each control's status, accept risks and flag issues. The contents can be downloaded as a report for auditors and stakeholders, with automated checks and owner-confirmed items marked separately. Anthropic states that Security Center is guidance based on its recommendations and does not guarantee compliance with laws or standards.

Parent-organization console (beta). For companies with multiple organizations, a parent console manages company-wide settings across Claude Enterprise and Claude Platform organizations: single sign-on (SSO), directory sync (SCIM), domain verification, groups and Compliance API keys. It adds two roles, parent organization owner and parent organization administrator. Multiple identity providers can be connected, which supports keeping an acquired company's existing IdP or migrating domain by domain.

Members, roles, settings and organization-level controls (IP allowlists, tenant restrictions, session duration) remain with each individual organization. Setup requires a child-organization administrator to propose the parent administrator and all organization administrators to approve. The setup cannot be reversed once completed, so staffing decisions must be made beforehand.

Access control and shadow IT (beta).

• One Admin API call signs a member out of all browsers and devices without removing them from the organization, and requires re-sign-in to apps such as Claude Code.
• Administrators can now manage IP allowlists directly. This previously required a procedure with Anthropic.
• Claude Desktop key settings can be applied from the admin console without MDM distribution.
• Claude Desktop on managed PCs can block use of personal or external-organization accounts.
• Owners can restrict which workspace folders users can use in Claude.

EDR integration (beta). Cowork runs shell commands and code in isolation on the local machine. Companies can now install their endpoint detection and response (EDR) sensor inside the virtual machine of the integrated Claude app. The VM verifies the EDR vendor's package signature at startup before installing. The sensor runs in detection mode, reporting activity without blocking it, and sends logs to the company's existing monitoring console.

EDR integration is in beta for macOS and Windows. It covers Claude Enterprise and Claude Team, and the Claude desktop app on Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry via MDM.

What this means

The generally available piece is the inference hooks, which let security teams extend existing DLP policies to Claude prompts and tool calls. The rest of the update is in beta, so administrators should expect these controls to change. Two constraints affect rollout planning: the parent console cannot be undone after setup, and the EDR sensor in Cowork only reports activity and does not block it. Security Center's report can document control status for auditors, but Anthropic says it does not establish compliance on its own.