xAI has published documentation on Grok Bot for teams and enterprises, detailing its architecture, administrative controls, and security model. This information is intended for team and organization administrators rolling out Grok Bot, as well as security reviewers and members seeking to understand the approval processes.
Grok Bot is described as a computer-use agent that operates applications, browsers, and development environments. It runs in Cursor's cloud, with each user's work executing on a dedicated cloud computer. Desktop and mobile applications serve as thin clients for chat, review, and approvals.
Key Points
- Grok Bot is included with Teams plans, while several settings are exclusive to Enterprise plans and are not visible to self-serve Teams.
- The security model for Grok Bot is built on four principles: per-user isolation, no access by default, human approval gates, and administrative control.
- Each user's work runs in a dedicated Firecracker microVM, providing hardware-level separation from other users.
- Within a single user's environment, all Bots share one computer, isolating personalities and workspaces rather than compute resources.
- Sensitive actions require user approval, which is evaluated by an independent review model called Auto Review.
- Enterprise-only controls include an organization-wide Enable Grok Bot switch, Network Controls, Team Setup, Action Recording, and computer management for organization admins.
- Audit logs, OpenTelemetry Export, the MCP allowlist, and SCIM are also features exclusive to Enterprise plans.
Context
According to xAI documentation, Grok Bot provides standing Bots for everyday work such as research, operations, documents, browsing, and automations. The documentation specifies that the grok-4.6 model is used. The security model, network policy, approvals, logging, and data handling details are further elaborated in the Grok Bot security and Grok Bot security FAQ documents. These resources address common questions from security reviews regarding isolation, network and egress, approvals, logging, and data handling.
Why It Matters
These details provide administrators and security teams with critical information for deploying and managing Grok Bot within their organizations. Understanding the plan-specific features and security architecture allows for informed decisions regarding access, data handling, and compliance, particularly for security-sensitive deployments.
What To Do
- Review the Grok Bot for teams and enterprises documentation to understand the full list of admin controls and their availability across different plans.
- Consult the Grok Bot security and Grok Bot security FAQ for detailed information on the security model, network policies, and data handling practices.
- Note that Network Controls and the organization-wide Enable Grok Bot switch are Enterprise-only features, impacting how self-serve Teams can manage network access and overall bot deployment.
- For security-sensitive deployments, contact your Cursor account team to discuss Enterprise plan enablement, egress ranges, residency commitments, and security review support.