Anthropic Opens a Claude Enterprise Security Center in Beta and Makes Inference Hooks Generally Available
Anthropic has added a security center, in beta, that compares an organization's Claude settings with Anthropic's recommended configuration and can export a report. Inference hooks — prompt hooks and pre-tool-call hooks that send prompts and tool calls to a customer's existing data-loss-prevention and security tools — are now generally available. Other additions, several of them still in beta, include a parent-level admin console at admin.claude.com, Admin API member sign-out, a self-serve IP allowlist, Claude Desktop controls aimed at shadow IT, and an endpoint-detection sensor that can run inside the local Cowork virtual machine in detection mode.
Anthropic is selling Claude Enterprise less as a better chatbot and more as something a security team can inspect and intercept. The company says a new security center, now in beta, compares an organization's Claude settings for chat, Cowork and Claude Code with Anthropic's recommended configuration, shows current versus recommended values, and lets admins record attestations and download a report. Separately, inference hooks are generally available: prompt hooks send prompts from supported Claude products to a customer's security tools, and pre-tool-call hooks check tool calls before they run.
The hooks are the sharper change. They let a company apply data-loss-prevention policies it already wrote to Claude traffic, rather than trusting Claude's own filters alone. Anthropic says they work across Claude, Claude Code, Cowork and other Enterprise products. The security center, by contrast, is explicitly "guidance based on Anthropic's recommendations, not a certification or a guarantee of compliance with any law or standard." Organizations with the Compliance API can read the same results through that API.
The release continues a run of Claude Enterprise admin features that already includes smart reports, in-app surveys and pooled group budgets and the unified Claude chat, Cowork and Enterprise workspace.
What else is shipping, and what is still beta#
A centralized management console at admin.claude.com, also in beta, sits above individual Claude Enterprise and Claude Platform organizations. Parent-level owners and admins can set company-wide identity controls — single sign-on, SCIM directory sync, domain verification, groups and Compliance API keys — once, and can connect more than one identity provider. Each child organization keeps its own members, roles and organization-level controls such as IP allowlisting and session duration. Setup cannot be undone, Anthropic says.
Access controls added in this round include signing a member out through a single Admin API call without removing them from the organization (the call ends browser and device sessions but not API keys, already-running cloud sessions or scheduled routines); a self-serve IP allowlist that no longer requires going through Anthropic; and Claude Desktop settings that can block personal or other outside Claude accounts on managed computers. Anthropic lists member sign-out, the self-serve allowlist, Desktop settings and centralized management as beta.
Admins can also run their organization's endpoint-detection and response sensor inside the isolated virtual machine that Cowork and unified Claude use for local shell and code on a user's computer. The sensor is installed after Claude verifies the vendor package's signature, reports activity to the customer's existing console and runs in detection mode — it does not block actions. The integration is in beta on macOS and Windows, covers the on-device virtual machine rather than cloud Cowork sessions, and Anthropic says it is also available for the Claude desktop app on Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry when deployed through device management.
Limits#
- Security center, centralized management, Admin API sign-out, the self-serve IP allowlist, Desktop controls and the EDR integration are in beta; inference hooks are the generally available piece.
- The security center is Anthropic's recommended configuration, not a compliance certification.
- Which security vendors the hooks and EDR integration support is left to Anthropic's help center; we did not audit that list.
- We have not tested the hooks against a third-party DLP product.