Anthropic adds security dashboard, DLP hooks and personal-account blocking to Claude Enterprise
Anthropic expanded Claude Enterprise's admin tools so security teams can see settings, inspect data flows and block personal logins on work PCs.
The short version
- Inference hooks, which send prompts and tool calls to existing DLP tools for inspection, are generally available.
- A Security Center page compares an organization's Claude settings with Anthropic's recommended setup (beta).
- Claude Desktop on managed PCs can block personal and outside-organization accounts (beta).
- EDR sensors can run inside Cowork's virtual machine in detection-only mode (beta, macOS and Windows).
- Anthropic says the Security Center does not guarantee compliance with any law or standard.
Why it matters
Staff pasting work material into a private AI login is a common worry for companies. These controls give administrators a way to see and limit it, but only if their security team already runs the matching tools. They are a vendor's governance answer, not a fix for any specific incident.
What Anthropic announced
As kn.itmedia.co.jp reported, Anthropic said on November 5, 2026 that it had expanded security and management features in Claude Enterprise, its business plan. The aim is to let security staff see how Claude is configured, what it does on employee devices, and how it handles data during work.
Only one piece is fully launched: the inference hooks. The rest is in beta, meaning still being tested and subject to change.
Seeing settings and checking data
The new Security Center is a single page inside an organization's settings. It lists security settings for chat, Cowork and Claude Code, shows each current value next to Anthropic's recommended one, and re-checks on every page load. Reports can be downloaded for auditors. Anthropic stresses it is guidance, not proof of legal compliance.
Inference hooks connect Claude to data loss prevention (DLP) tools, which stop sensitive data leaving a company. A prompt hook inspects what a user types. A pre-tool-call hook checks before Claude uses a tool. Existing policies can be reused as they are.
Closing the personal-account gap
For Claude Desktop, admins can apply key settings from the admin page without device-management software. On managed PCs they can block personal accounts and accounts from other organizations, which targets shadow IT: tools staff use without approval. Owners can also limit which workspace folders Claude can use.
Admins can now force sign-out from every browser and device with one Admin API call, without removing the member. They can also edit IP allowlists themselves instead of asking Anthropic.
EDR and multi-organization control
Cowork runs shell commands and code in isolation on a local machine. Companies can now install their endpoint detection and response (EDR) sensor, security software that watches devices, inside that virtual machine. It runs in detection mode, so it reports activity to the existing monitoring console but does not block anything.
Large companies get a parent-organization console to manage single sign-on, directory sync and domain verification across child organizations. Setup cannot be undone, so choose the responsible people first.
What to watch
- Which features leave beta, and whether pricing or plan limits apply.
- How well the hooks and EDR sensors work with the security products companies already run.
Try it
If you administer Claude Enterprise, check whether your team already runs a DLP or EDR product, then test the personal-account block on one managed PC first.
Sources and links
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by kn.itmedia.co.jp · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
