rencore.com (opens in a new tab) Start a trial (opens in a new tab)

Policies for ChatGPT Enterprise

Last updated Preview Policies ChatGPT Enterprise

AI summary

ChatGPT Enterprise policy templates help you identify governance issues across seven categories: security, guest access, oversharing, inactive accounts, unused resources, licensing, and operations. Each template checks specific objects and flags violations with severity ratings; recommended templates guide you to prioritize which checks to run first.

These are the policy templates for ChatGPT Enterprise, grouped by what they look for. Each template checks one kind of object and flags every item that breaks the rule. The severity shows how serious a violation is; a Recommended pill marks the templates Rencore recommends you start with.

See the ChatGPT Enterprise inventory reference for the objects these templates work on.

Security 5

Risky settings, weak credentials and access that is wider than it should be.

PolicyDescriptionSeverityChecks
Agent runs without a human and reaches external systems RecommendedDetects published agents that ran from a schedule, webhook or API call in the last 30 days and can reach a connector.CriticalChatGPT agent
Pending invitation grants owner or admin rights RecommendedDetects pending invitations that would grant the owner or admin role on acceptance.HighChatGPT invitation
ChatGPT login bypassed SSO RecommendedDetects users who signed in with a password rather than through the identity provider in the last 30 days.HighChatGPT user
ChatGPT user with no Entra ID identity RecommendedDetects ChatGPT users who do not match any Entra ID user by email.HighChatGPT user
ChatGPT workspace with too many admins RecommendedDetects workspaces where more than five users hold the admin role.MediumChatGPT Workspace

Guests and external access 5

Guests, external users and content shared outside your organization.

PolicyDescriptionSeverityChecks
Shared Codex environment with repository access RecommendedDetects Codex sandboxes that are shared and can reach source repositories.HighChatGPT Codex Environment
Invitation sent to an external domain RecommendedDetects invitations to addresses that match no Entra ID user or match a guest.MediumChatGPT invitation
Custom GPT calls an external domain without authentication RecommendedDetects custom actions that call an external host with no authentication configured.HighChatGPT custom GPT
Plugin listed to the whole workspace bundles an MCP server RecommendedDetects discoverable plugins that install an MCP server any member can reach.HighChatGPT plugin
Over-shared project with external data sources RecommendedDetects ChatGPT projects that are broadly shared and also pull from an external system.CriticalChatGPT project

Oversharing 3

Content that more people can open than need to.

PolicyDescriptionSeverityChecks
Publicly reachable custom GPTs RecommendedDetects custom GPTs reachable outside the workspace by link or through the GPT store.HighChatGPT custom GPT
Publicly reachable ChatGPT projects RecommendedDetects ChatGPT projects reachable outside the workspace or by every member.HighChatGPT project
Workspace-wide custom GPTs with editable configuration RecommendedDetects custom GPTs every workspace member can reach and edit.MediumChatGPT custom GPT

Leavers 4

What people who left, or whose accounts are disabled, still own or hold.

PolicyDescriptionSeverityChecks
Inactive user still holding objects RecommendedDetects inactive workspace users who still own GPTs, projects, agents, Codex environments or library files.HighChatGPT user
Manually managed ChatGPT user RecommendedDetects workspace users who are not provisioned through SCIM.MediumChatGPT user
Custom GPT with an inactive owner RecommendedDetects custom GPTs whose owner is inactive or no longer in the workspace.MediumChatGPT custom GPT
ChatGPT project with an inactive owner RecommendedDetects ChatGPT projects whose owner is inactive or no longer in the workspace.MediumChatGPT project

Clean-up 5

Inactive, empty, ownerless or outdated objects you can archive or delete.

PolicyDescriptionSeverityChecks
Idle Codex machine still pairedDetects developer machines not seen for 30 days that are still paired for remote control.LowChatGPT Codex machine
Stale custom GPTDetects custom GPTs whose configuration has not changed in 90 days.LowChatGPT custom GPT
Stale invitationDetects invitations still pending more than 30 days after they were sent.LowChatGPT invitation
Stale ChatGPT projectDetects ChatGPT projects whose configuration has not changed in 90 days.LowChatGPT project
Unused Codex environment with repository access RecommendedDetects Codex sandboxes that can reach repositories but have never run a task.MediumChatGPT Codex Environment

Licenses and costs 5

Licenses, spend and storage that cost money without being used.

PolicyDescriptionSeverityChecks
Codex seat with no Codex usage RecommendedDetects users on a Codex seat who have not used Codex in 30 days.LowChatGPT user
Group without a default spend limitDetects groups with no default cap, whose members fall through to the workspace default.LowChatGPT group
Idle ChatGPT seat RecommendedDetects users with no usage this month or last month.LowChatGPT user
ChatGPT user with no spend cap RecommendedDetects users whose effective spend limit is unlimited.MediumChatGPT user
ChatGPT user over 80% of their spend cap RecommendedDetects users whose month-to-date usage has passed 80% of their effective cap.MediumChatGPT user

Operations 3

Day-to-day hygiene: naming, ownership, settings and configuration.

PolicyDescriptionSeverityChecks
Connector scope failing to sync RecommendedDetects external data sources attached to a project whose last sync failed.MediumChatGPT connector scope
Service account credential expiring within 30 days RecommendedDetects active service account credentials that lapse in the next 30 days.MediumChatGPT service account credential
Shared custom GPTs affected by the GPT retirement RecommendedDetects custom GPTs shared beyond their owner, which stop working when OpenAI retires custom GPTs.HighChatGPT custom GPT

Frequently asked questions

How many built-in policies does Rencore Governance provide for ChatGPT Enterprise?

30 built-in policies, such as Agent runs without a human and reaches external systems, Codex seat with no Codex usage and 28 more.

How are the ChatGPT Enterprise policies grouped?

By category: Security, Guests and external access, Oversharing, Leavers, Clean-up, Licenses and costs and Operations.

Which ChatGPT Enterprise policies are recommended?

Agent runs without a human and reaches external systems, Codex seat with no Codex usage, Connector scope failing to sync and 22 more.

Which module includes the ChatGPT Enterprise connector?

The ChatGPT Enterprise connector is part of the AI & Agents module of Rencore Governance, an add-on module on the Premium and Enterprise Plans. The connector is in preview.