Security
Anthropic adds a Security Center and generally available inference hooks to Claude Enterprise; a company's own security tools, not Anthropic, scan the prompts and tool calls
Claude Enterprise gets a Security Center (beta) and generally available inference hooks that send prompts and tool calls to a company's own security tools.
- Coding agents
- #Claude Enterprise
- #Anthropic
- #Security Center
- #inference hooks
- #DLP
- #tool calls
- #enterprise security
On November 5, 2026, Anthropic announced security and admin updates for Claude Enterprise. Two pieces matter most: a Security Center, in beta, that compares an organization's Claude settings with Anthropic's recommended configuration, and inference hooks, now generally available, that send prompts and tool calls to the company's own security tools for a check. Disclosure: Anthropic makes Claude, which helps draft articles on this site. We read Anthropic's page through search excerpts and the rest through press and vendor pages.
The essentials
- Two different things: the Security Center is a settings audit; inference hooks are a control applied while people work. Only the hooks are generally available.
- Who scans: the company's existing security or data loss prevention (DLP) tool does. Anthropic sends it the prompt or the proposed tool call; Anthropic itself is not described as doing the scanning.
- Two hook types: prompt hooks check prompts from supported Claude products; pre-tool-call hooks check tool calls before they run.
- Not a certificate: Anthropic says the Security Center does not certify compliance or guarantee adherence to legal or industry standards, per a press relay.
What does the Security Center actually do?
It is a new page in Organization settings that compares settings for chat, Cowork and Claude Code with Anthropic's recommended configuration, so a team can see what to change and where. It checks the configuration each time the page loads, and results can be exported as an audit report. It is in beta, and Anthropic lists some exceptions during the beta in its Help Center. An update also covers a parent-organization console, in beta, for identity settings such as SSO and SCIM across Claude Enterprise and Platform organizations.
How do the hooks work, and what are their limits?
An admin names a security server, and Anthropic calls it over HTTPS before a prompt or tool call goes ahead. One vendor, Speakeasy, describes the flow as allow, deny or review, and says its own integration does not rewrite prompts, does not block tool or assistant responses, and does not inspect image-only attachments because Anthropic sends metadata and extracted text rather than raw images. Those are one vendor's statements about its product, not necessarily Anthropic's general rules. Separately, the Compliance API, reported in September, feeds Claude Enterprise activity to more than 100 security vendors for monitoring; the hooks are described as acting as people work.
What is not established
- Exactly which Claude products and tool paths hooks cover: sources list them slightly differently.
- What happens if a customer's security server is slow or unreachable: not stated in the sources we read.
- How well the Security Center's recommended configuration fits a given company: it is Anthropic's own benchmark, not an outside standard.
- Independent testing of hook coverage or bypass resistance: none found.
Summary table
| Point | What is stated as of November 6, 2026 | Source |
|---|---|---|
| Date | November 5, 2026 | Claude, AI Weekly |
| Security Center | Beta; compares chat, Cowork and Claude Code settings with Anthropic's recommendation | Claude, AI Weekly |
| Inference hooks | Generally available; prompt hooks and pre-tool-call hooks | Claude, Superpower Daily |
| Who inspects | The customer's own security or DLP tool | Claude, Speakeasy |
| Certification | Security Center does not certify compliance | Superpower Daily |
| Plan | Claude Enterprise, with some beta exceptions | Claude |
FAQ
Does Anthropic read my prompts to scan them?
The sources describe prompts being sent to the customer's own security tool for a check. They do not describe Anthropic running the DLP scan.
Is the Security Center a compliance certificate?
No. Anthropic says it does not certify compliance, and it is still in beta.
Which plan gets this?
Claude Enterprise, with some exceptions during the beta listed in Anthropic's Help Center.
Method and limits
Sources checked on November 6, 2026. Anthropic's page was read through search excerpts, not in full; Speakeasy is a vendor describing its own product. We did not test the hooks. This is not security or compliance advice.
Sources
- Claude: More visibility and control over Claude Enterprise for security teams (verified 2026-11-06)
- AI Weekly: Anthropic Ships Claude Security Center, Inference Hooks GA (verified 2026-11-06)
- Superpower Daily: Anthropic adds security checks and company-wide controls to Claude Enterprise (verified 2026-11-06)
- Speakeasy: Enterprise security for the whole Anthropic suite (a vendor describing its own hook integration) (verified 2026-11-06)
- Help Net Security: Security tools can now scan Claude Enterprise chats and uploads for sensitive data (verified 2026-11-06)
Spotted an error in this article? Report it · Our corrections policy
