Claude Enterprise adds security oversight and access controls

What changed

Anthropic’s update adds configuration oversight and access controls for Enterprise organisations. The security center is available to organisation Owners and the Primary Owner; the Admin role alone cannot access it. Its security center compares organisation settings with Anthropic recommendations. It is guidance, not certification. Centralised administration and the security center are in beta; prompt and pre-tool-call inference hooks are generally available for supported products.

For teams already using Claude with campaign material or client information, the practical question is which controls cover their actual workflow. A central console can help document a review, but it does not replace checking the organisation’s approved uses, access rules and responsibilities. Record which products and user groups a control applies to before treating it as protection for every session.

One important limit concerns the beta Admin API sign-out capability. Signing a member out does not revoke API keys, stop running cloud sessions or cancel scheduled routines. An offboarding procedure therefore needs separate checks for each route of access. For example, a team retiring a contractor’s account should verify both interactive access and any automation that account helped configure.

The announcement also discusses recently introduced endpoint monitoring. That should not be treated as a newly launched prevention system across all Claude environments. Teams should match their control inventory to the current product documentation rather than assume every capability in a roundup shares the same release date.

Why it matters for advertisers

Agency and in-house teams can use this announcement to review existing AI governance. The useful distinction is between a control being documented, enabled for the organisation and verified in the workflow that handles client data.