Qubit Docs العربية

Model governance

This page explains how the AI features in Qubit are scoped to a workspace, how you control what they can do, and what Qubit keeps after a run.

Workspace scope

The Assistant works inside the workspace you are in. It uses that workspace's data connections, data model, documents, and organization context, and it queries only the data your access allows. See What the AI can and cannot access.

Two personal settings follow you across workspaces: your agent instructions and your capability switches. They describe your role and what the Assistant may change for you. See Personal agent settings.

Controls you have

  • Data access: data permissions and row restrictions limit what the Assistant can query, exactly as they limit what you can query. See Access Control.
  • Actions: each person decides whether the Assistant can create or edit dashboards, reports, metrics and the semantic model, insights, scheduled tasks, and data source descriptions on their behalf. Sharing, activating schedules, deleting, and connection credentials stay under your control in the product.
  • Context: owners and admins decide what the Assistant knows about the company in the workspace Agent tab. See Workspace AI agent.

What Qubit keeps

  • Conversations: your chats with the Assistant are saved so you can return to them.
  • AI run records: while audit recording is on, each AI run is recorded in the workspace audit log with its request, answer, model, SQL statements, tool calls, token usage, and the person's access at the time. Records are kept for the retention period the workspace sets, from 30 to 2,555 days. See Audit Logs.

Owners and admins can export these records or forward them to their own security monitoring system.