ChatGPT dots for Enterprise admins: setup, controls and open questions

If you run a ChatGPT workspace, you’re probably looking for the admin toggle, the policy controls and the audit story for dots. OpenAI’s learn docs include an admin guide, Manage dots permissions and capabilities, that covers most of this. This page sums up that guide, adds what the help center and privacy FAQ say, lists what is still unannounced, and ends with a pilot checklist that is our own suggestion.

Who can get dots

Users must be 18 or older. The texting beta is limited to Pro users in the US, so it isn’t available in Business or Enterprise workspaces.

Turn dots on: Permissions & roles

OpenAI’s admin guide puts all of this under Workspace settings > Permissions & roles, using workspace defaults or custom roles assigned to groups. There are four dots permissions:

Permission	What it allows
Use dots (shown as Use dots (Beta))	Members can use dots
Add dots to Slack	Members can add their dot to Slack
Allow local computer access	Dots can use a member’s local computer. This is a separate admin opt-in.
Use custom rules for dots	Members can add or edit rules for their dots

To enable dots for everyone, turn on Use dots (Beta) under Permissions > Workspace default > Workspace capabilities. To enable it for a pilot group only, grant it through a custom role instead.

The same settings page has cloud computer capabilities: Cloud browser use, Cloud network access and Cloud computer use. These decide whether a dot’s cloud computer can browse websites, reach online services and work in apps. Use password manager sits under Workspace capabilities. OpenAI says to review how these settings apply across both dots and Work Cloud before changing them.

Apps, Slack and Teams

Apps. Dots use the workspace’s existing ChatGPT app connections. You set app access and allowed actions in Plugin controls, the same place as for ChatGPT. The privacy FAQ adds that plugin permissions are shared across dots, ChatGPT, ChatGPT Work and Codex, so there is no separate dots-only plugin layer. See connected apps.

Slack needs three parties:

1. A ChatGPT workspace owner enables Use dots and Add dots to Slack for the intended members.
2. A Slack workspace owner or app manager makes sure the ChatGPT app is installed in Slack and approves app requests if required.
3. Each member connects their dot to Slack from its profile.

In Slack, OpenAI says “Only the owner can direct their dot through a Slack direct message or supported channel mention. Messages from other people do not start work.” Anyone who can see the conversation can see what the dot posts there.

Teams. The user docs say you can message your dot in Slack and Microsoft Teams. The admin guide is narrower: “Microsoft Teams access is limited to an invite-only alpha.” For an Enterprise rollout, plan on Slack.

Local computers and data residency

A dot’s cloud computer is separate from a member’s laptop. Local sign-ins, VPN access and device policies don’t carry over to it. Local work requires a connected computer that is online with ChatGPT open and access enabled.

One policy interaction matters: “If any cloud policy has enforce_residency enabled, local computer access is unavailable for dots. Work Cloud remains available.”

What the admin controls do and don’t cover

Area	Admin control	Caveat
Access	Grant or revoke dots through workspace permissions	Revoking dots doesn’t disconnect apps or sign out website sessions. Review those separately.
Cloud and local computers	Cloud browser use, Cloud network access, Cloud computer use, Use password manager, local computer access	Local access is a separate opt-in and is unavailable under enforce_residency
Messaging	Enable or disable Slack participation	Teams is an invite-only alpha
Connected apps	Available apps and allowed actions in Plugin controls	Shared with ChatGPT, ChatGPT Work and Codex
Custom rules	Turn member rules on or off	When off, saved rules don’t apply, but default rules and explicit user instructions still do
Models	None	Enterprise model controls and defaults don’t apply to dots
Usage and adoption	Analytics API for available adoption metrics	Usage terms after the first month aren’t published yet
Audit	Compliance API records of user messages and dots’ replies	OpenAI says to confirm record coverage before relying on it for an audit

Turning custom rules off doesn’t remove the safety floor. In OpenAI’s words, default rules “still apply when custom rules are off.” Turning custom rules off also doesn’t mean every action needs confirmation, so review capability and app approval settings as well. See custom rules.

Points that matter for governance

Training. OpenAI’s privacy FAQ says Business, Enterprise and Edu data is not used for training by default. It also says human review may occur in limited circumstances, including safety-related cases.

Deleting a dot doesn’t delete everything. Deleting a dot removes its own context. Files, Codex threads and ChatGPT conversations it created are stored separately and are not deleted, and ChatGPT memories remain. Disconnecting a plugin stops new access but doesn’t remove what the dot already holds. OpenAI’s admin guide suggests reviewing saved memories and reset options when handling sensitive data or offboarding. See memory and reset.

Built-in guardrails apply everywhere. Changing a password or transferring money always requires the user to take over. Auto-review checks certain actions, such as sending an email, before they happen. The admin guide says “Drafting a message does not authorize sending it.” Prompt-injection protections reduce the risk but, in OpenAI’s words, “do not eliminate” it.

What’s still unannounced

• Usage terms after the first month. The help center now describes an allowance for deeper work with extended limits for the first month. See pricing and usage.
• Prices for adding more dots or scaling one up.
• When Teams leaves its invite-only alpha.
• Availability and pricing for specialist dots.

Specialist dots

The announcement also previews specialist dots, which have their own identity and IT-provisioned hardware. OpenAI says they are “starting with focused enterprise pilots” and that a Microsoft Agent 365 integration is in progress. There is no public sign-up path or availability date in the sources we track. The primary dot described on the rest of this site is separate from this preview.

A rollout checklist for admins

Step	What to decide	Why
1. Pilot group	A custom role with Use dots for a small group of users who are 18 or older and comfortable reviewing agent work	You can grant dots through a role instead of the workspace default, and Enterprise access is a beta.
2. Cloud computer scope	Which of Cloud browser use, Cloud network access and Cloud computer use the pilot needs	These settings are shared with Work Cloud, so check both before you change them.
3. Plugin scope	Which apps and actions are allowed in Plugin controls	Permissions are shared with ChatGPT, Work and Codex, so a change affects all four.
4. Custom rules	Whether members may write rules, plus a written default such as asking before any external email	Turning rules off makes saved rules stop applying, while default rules still apply.
5. Local computer access	Whether to opt in at all, and whether any cloud policy uses enforce_residency	Local access is a separate opt-in, and residency enforcement disables it for dots.
6. Channels	Slack permission plus the ChatGPT app in your Slack workspace	Teams is an invite-only alpha, and the ChatGPT permission alone doesn’t install anything in Slack.
7. Oversight	A weekly Activity View review with pilot users, plus a Compliance API coverage check	Activity View is the user-side record, and OpenAI says to confirm Compliance API coverage before an audit.
8. Offboarding	Reset dots and disconnect apps when people leave	Disconnecting an app doesn’t erase what a dot already holds.

Related reading

For plan and region detail, see availability. For how individual users control what a dot can do, start with custom rules and connected apps.