Anthropic adds a security center and generally available inference hooks to Claude Enterprise

What Anthropic announced

Anthropic published More visibility and control over Claude Enterprise for security teams on 5 November 2026. The update adds a security center, makes inference hooks generally available, introduces centralized management across several Claude organizations and adds new access controls. Anthropic frames it around three needs of security teams: seeing how Claude is configured, what it does on company devices and how data is handled as people work.

The security center is a new page in Organization settings. For each control across chat, Cowork and Claude Code it shows the current value, Anthropic's recommended value where there is one, and a link to the setting. It flags new or recently changed settings, rechecks the configuration each time the page loads and can be exported as a report for auditors. Admins can record attestations, accept risks and flag issues together. Anthropic states that the security center is guidance, not a certification or a guarantee of compliance with any law or standard.

Inference hooks bring existing DLP rules to Claude

The change with the most practical weight is that inference hooks are now generally available. Prompt hooks send prompts from supported Claude products to a company's security tool for a check. Pre-tool-call hooks check a tool call before Claude runs it. Anthropic says the hooks work with security vendors, so a team can keep the data loss prevention policies it has already written and apply them to Claude instead of building separate rules.

That matters because agentic products act, not only answer. A pre-tool-call check gives a security team a point at which a command, file write or connector call can be inspected before it happens. For organisations that have held back Cowork or Claude Code because their DLP stack could not see inside them, the hooks are the missing piece.

One console, sign-outs and a block on personal accounts

Centralized management, in beta, gives a parent organization its own console at admin.claude.com. Admins set single sign-on, SCIM directory sync, domain verification, groups and Compliance API keys once for every Claude Enterprise and Claude Platform organization under the parent. Companies can connect more than one identity provider, which Anthropic says lets an acquired company keep its own. Setup needs approval from admins of each organization and cannot be undone.

New access controls, also in beta, let admins sign a member out of every browser and device with one Admin API call without removing them. Anthropic notes that this does not end API keys, running cloud sessions or scheduled routines. Enterprise admins can now manage their IP allowlist without going through Anthropic, and can block personal and other outside Claude accounts in Claude Desktop on managed computers, a direct answer to shadow IT.

Endpoint visibility and what to watch

Anthropic also points to its recently announced EDR integration. When Cowork and unified Claude run locally, they execute commands in an isolated virtual machine on the user's computer. Admins can now run their EDR vendor's sensor inside that machine through device management on macOS and Windows. Claude verifies the sensor package's signature each time the machine starts, and sensors run in detection mode: they report activity but do not block it. Cloud Cowork sessions are not covered.

The update lands in a week when AI tools kept moving up security teams' lists, including the Claude Code account email leak reported on 6 November. For a security team, the combination of hooks, a settings audit page and endpoint sensors is what turns an AI assistant from an unmanaged app into something that can pass an internal review. Open questions are which DLP and EDR vendors are supported, listed only in the Help Center, and when the beta features become generally available. Anthropic has not given dates. Related reporting on AI security tools is in our cybersecurity section.

Questions people ask

What are Claude inference hooks?

They send prompts (prompt hooks) or requested tool actions (pre-tool-call hooks) from supported Claude products to a company's own security tool for a check, so existing DLP policies apply. They are generally available for Claude Enterprise.

Is the Claude Enterprise security center a compliance certification?

No. Anthropic says it compares settings with its recommended configuration and is guidance, not a certification or guarantee of compliance with any law or standard.

Can admins stop employees using personal Claude accounts?

Yes, in beta. Admins can block personal and other outside Claude accounts in Claude Desktop on managed computers.

Sources

1. More visibility and control over Claude Enterprise for security teams (Anthropic, 5 Nov 2026)