Claude Enterprise adds a security center, DLP hooks and EDR support for security teams

On 5 November 2026, Anthropic announced a set of security and admin updates for Claude Enterprise, aimed at the teams that have to approve and monitor AI tools inside a company. One feature, inference hooks, is generally available. The rest are in beta.

What's generally available: inference hooks

Inference hooks connect Claude to the data loss prevention (DLP) and security tools a company already runs. DLP tools watch for sensitive data, such as customer records or source code, leaving the company.

• Prompt hooks send prompts from supported Claude products to your security tool to be checked.
• Pre-tool-call hooks check an action Claude is about to take, such as calling a connected app, before it runs.

Each check goes to a server your company hosts, which answers allow or deny before Claude continues. Anthropic's Help Center says uploaded file text and tool responses can be checked too. Because the check runs in Claude's own infrastructure, nothing has to be installed on staff computers, and existing company policies apply to Claude without a separate rulebook. Owners turn hooks on in Organization settings.

What's in beta

• Security center. A page that compares your settings for chat, Cowork and Claude Code with Anthropic's recommended configuration. It shows current and recommended values, flags new or changed settings, lets admins record accepted risks, and exports a report for auditors. Anthropic is clear this is guidance, not a certification of compliance.
• Centralised management. One admin console for a parent organisation covering several Claude Enterprise and Claude Platform organisations, with shared single sign-on, user provisioning, domain verification and groups. Each organisation keeps its own members and settings. Setup can't be undone.
• Sign a member out everywhere. One Admin API call ends a person's sessions on all browsers and devices without removing them. It doesn't revoke API keys, cloud sessions already running or scheduled tasks.
• Self-serve IP allowlist. Admins can limit access to most Claude products to approved company IP addresses themselves.
• Claude Desktop controls. Admins can set common desktop settings without device-management pushes, block personal Claude accounts on managed computers to cut shadow IT, and limit which folders people can use with Claude.
• EDR inside Cowork. Your endpoint detection and response (EDR) sensor, the security software that watches computers for threats, can run inside the virtual machine Cowork uses on a person's computer and report to your existing security console. It detects and reports but doesn't block, and covers local sessions, not cloud ones.

Who should care

• Security and IT teams asked to approve Claude for wider use, who need it to fit existing DLP, EDR and identity tooling.
• Groups with several Claude organisations, for example after an acquisition, who want one place to manage sign-in and admins.
• Compliance leads who need evidence for auditors of how AI tools are configured.

Limits and caveats

• Most features are for Claude Enterprise. Anthropic lists beta exceptions in its Help Center. The EDR feature also covers Claude Team.
• Everything except inference hooks is beta, so expect changes before relying on it for audits.
• Hooks need a server you run that can answer allow or deny quickly, usually via your security tool. EDR works only with supported vendors, listed in Anthropic's Help Center.
• A security center score is a configuration check, not proof of compliance with any law or standard.

What to do this week

• If you run Claude Enterprise, open the security center and review every control that differs from the recommendation. Record why you accept any gaps.
• Ask your DLP vendor whether it can act as the server for Claude inference hooks, and start with pre-tool-call hooks on connectors that touch customer data.
• Add the member sign-out API call to your staff offboarding checklist, alongside revoking API keys, which it doesn't cover.
• If staff use personal Claude accounts on work machines, plan the switch before turning on the block.

The DevDrift builds AI automation for regulated teams with role-based access and audit trails, and can help fit Claude into your existing security controls.

Sources

• Anthropic: More visibility and control over Claude Enterprise for security teams
• Claude Help Center: Inference hooks overview