Anthropic around 5 November 2026 shipped Claude Enterprise security updates led by generally available inference hooks that send prompts and tool calls to an organisation’s own data-loss-prevention and security tools before Claude continues. The package also adds a Security Center in beta, a parent admin console at admin.claude.com, stronger access controls, and an endpoint detection and response sensor inside the local Cowork and unified Claude virtual machine.

The company detailed the rollout in its Claude Enterprise security updates article. For Indian CIOs already rolling out Claude for coding and knowledge work, inference hooks are the clearest kill-switch-style control yet on what staff can send into the model and what tools it may call.

What Claude inference hooks deliver

Inference hooks are now generally available on Claude Enterprise. Prompt hooks forward prompts from supported Claude products to a customer-hosted security endpoint for an allow-or-deny decision before generation starts. Pre-tool-call hooks check tool calls the same way before tools run, including calls made through skills, plugins and connected tools.

Anthropic’s inference hooks overview says the check runs inside Claude’s infrastructure rather than on each employee device, so one setup covers Claude chat, Claude Code, Cowork and other Enterprise products. Common uses include data loss prevention, transcript archival and organisation-specific policy enforcement. Owners and Primary Owners turn hooks on and manage them.

• Prompt hooks and pre-tool-call hooks are generally available for Claude Enterprise
• Security Center (beta) compares chat, Cowork and Claude Code settings with Anthropic’s recommended configuration
• Centralized management (beta) at admin.claude.com covers SSO, SCIM, domain verification, groups and Compliance API keys

Security Center, admin console and access controls

The Security Center shows current versus recommended values for each control, flags new or recently changed settings, and lets admins download an auditor report. Attestations give owners a shared workspace to record control state, accept risks and flag issues. Anthropic stresses the page is guidance, not a certification of legal compliance.

Centralized management lets a parent organisation set identity and company-wide settings once across Claude Enterprise and Claude Platform organisations under that parent. New access controls allow Admin API member sign-out without removal from the organisation, self-serve IP allowlists, blocking personal and outside Claude accounts in Claude Desktop on managed computers, and limiting which workspace folders Claude may use.

EDR inside Cowork and why Indian CISOs care

When Cowork and unified Claude run locally, they execute shell and code in an isolated virtual machine. Admins can now run their organisation’s EDR sensor inside that VM so activity reports to the same console the security team already monitors. Sensors run in detection mode: they report and do not block. Setup is through device management on macOS and Windows for Claude Team and Enterprise, including desktop apps on Bedrock, Vertex AI and Microsoft Foundry via MDM.

The timing matters for Indian enterprises. CERT-In’s AI security audit rules already push six-hour reporting windows, while the agency has obtained Anthropic Mythos access for defensive evaluation. Partners such as Straive’s Anthropic enterprise AI programme are scaling Claude in regulated workflows, and TCS’s enterprise AI agents study found only about 10% of firms let agents act alone. Inference hooks sit beside Salesforce Agent Fabric’s kill switch as controls CIOs can demand before wider Claude rollout.

Key takeaways

• Claude Enterprise inference hooks (prompt and pre-tool-call) became generally available around 5 November 2026 for DLP and policy checks before inference and tool use.
• Security Center, admin.claude.com centralized management, Admin API sign-out, self-serve IP allowlists and Desktop shadow-IT blocks are in beta.
• EDR sensors can run in detection mode inside the local Cowork and unified Claude VM, reporting to the organisation’s existing security console.

FAQ

What are Claude inference hooks?

Inference hooks send prompts and tool calls from Claude Enterprise products to a security endpoint you host; Claude continues only after that endpoint allows the request, supporting DLP and custom policy enforcement.

Are all the new Claude Enterprise security features generally available?

No. Inference hooks are generally available; Security Center, centralized management at admin.claude.com, Admin API member sign-out, self-serve IP allowlisting and EDR in Cowork remain in beta per Anthropic’s Help Center.

Does the EDR sensor block activity inside the Claude VM?

Anthropic says supported EDR sensors run in detection mode: they report virtual-machine activity to your security console and do not block it.

Featured image: Anthropic CEO Dario Amodei at TechCrunch Disrupt 2023. Photo: Kimberly White/Getty Images for TechCrunch via Wikimedia Commons (CC BY 2.0)

Be a part of Elets Collaborative Initiatives. Join Us for Upcoming Events and explore business opportunities. Like us on Facebook , connect with us on LinkedIn and follow us on Twitter.

"Exciting news! Elets technomedia is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest insights!" Click here!