Start free trial
How-to 5 steps 4 min read Updated 11 November 2026

How to govern ChatGPT Enterprise and custom GPTs

Custom GPTs, projects and connectors multiply fast. Inventory them, see who each is shared with and join users to your Microsoft identity.

Torsten Mandelkow Works with
How-to

What is ChatGPT Enterprise governance?

Governing ChatGPT Enterprise means knowing which custom GPTs, projects, connectors and seats exist in your workspace, who each is shared with, what data it can reach and who owns it. It uses metadata and a join to your Microsoft identity, without reading conversations. Coverage in Rencore is in Preview.

In short

  • ChatGPT Enterprise keeps its own users, GPTs and projects, outside the Microsoft 365 policies you already run.
  • The first risks are sharing and reach: GPTs open to the whole workspace and projects wired to company data.
  • Govern from metadata. You do not need to read conversations to answer who owns what and who can open it.
  • Join every user to your identity provider, so seats and spend read by department.

Step by step

Five steps bring ChatGPT Enterprise under the same operating model as the rest of your digital workplace.

  1. Confirm the plan and connect the workspace

    Check that the workspace is ChatGPT Enterprise, since ChatGPT Business is outside OpenAI's Compliance Platform. The workspace owner creates the key for the connection, one key and one connection per workspace.

    Tip: Ask the workspace owner to create the key rather than sharing an admin login.

  2. Inventory GPTs, projects and connectors

    List every custom GPT, project, connector, workspace agent and service account with its owner and configuration. Include the connector scopes, such as the SharePoint sites and repositories a project can reach.

  3. See who each one is shared with

    For each GPT and project, record who it is shared with: individuals, groups or the whole workspace. Flag anything shared workspace-wide and anything that reads a company data source through a connector.

  4. Join users to your Microsoft identity

    Relate every ChatGPT Enterprise user to their Entra ID account, with department and country. This is how you spot users who hold both a ChatGPT Enterprise seat and a Copilot licence, and seats held by people who have left.

  5. Isolate before you delete

    Where a GPT or project is shared too widely, remove the sharing and keep it with its owner, then ask whether others still need it. Delete only what nobody claims, and keep a record of each step.

Why ChatGPT Enterprise needs its own governance

Employees build custom GPTs and projects the way they once built Teams and Power Apps: quickly, for a real need and with nothing recording who owns the result. A project connector can wire SharePoint, Slack, Google Drive or Linear into a project. A GPT shared with the whole workspace reaches everyone in it. Each is a small decision, and nobody sees them added up.

The workspace also sits beside Microsoft 365 with a separate user list. Without a join to Entra ID, a seat cannot be read by department or country, and nobody can check it against the Copilot licence the same person holds. OpenAI's own compliance log window is 30 days, which is short for any question about last quarter.

Connectors in this How-to

Each connector page in the Rencore docs lists the objects, policies, reports and actions behind these steps.

Common questions about governing ChatGPT Enterprise

Which ChatGPT plan can be governed this way?
A ChatGPT Enterprise workspace. OpenAI excludes ChatGPT Business from its Compliance Platform, so the controls described here do not exist there, and the OpenAI API platform, the developer side of OpenAI, is a different product with its own governance. Check which plan each team really uses before you promise coverage, because plans spread through expense reports.
Can IT read our people's ChatGPT conversations?
It should not need to. Governance runs on metadata: who owns a GPT or project, who it is shared with, which connectors and repositories it can reach, and the activity that shows who did what and when. Message text and file content are not required for that, and keeping them out of the governance record is a privacy decision you can state to employees and works councils plainly.
What is the biggest risk in custom GPTs?
Sharing and reach. A custom GPT shared with the whole workspace, or a project with a connector into a SharePoint site, exposes whatever that source holds to everyone who can open it. The same questions apply as for a Team or a site: who created it, who has access, what data it reaches and who cleans it up. Without an inventory, nobody can answer them.
Do we need this if we already use Microsoft Purview?
Purview governs data in Microsoft 365, and it is where your labels and data protection policies live. A ChatGPT Enterprise workspace keeps its own users, GPTs and projects outside that view, so a department can hold both a ChatGPT seat and a Copilot licence without anyone seeing the overlap. The two work together: Purview protects the data, an inventory shows which AI workspaces touch it.
How does Rencore help govern ChatGPT Enterprise?
Rencore Governance (Preview coverage for ChatGPT Enterprise) inventories custom GPTs, projects, connectors, agents and seats, shows who each GPT or project is shared with, and joins every user to their Microsoft Entra ID identity, from metadata only. Its automations can isolate a GPT or project so only the owner keeps it, before anything is deleted. The workspace owner creates the key Rencore connects with.

Last updated 11 November 2026

See ChatGPT Enterprise next to Microsoft 365

We will show custom GPTs, projects and seats joined to Entra ID in a short demo of the ChatGPT Enterprise Preview.

Rencore newsletter

Subscribe to our newsletter

Get the latest insights on governing Microsoft 365, Copilot and AI agents, delivered to your inbox.

Loading form