OpenAI dots are generally available to eligible Enterprise, Edu and Healthcare workspaces as of 4 November 2026, and they switch on by default once the admin preview ends on 18 November. If you run a workspace and haven't decided yet, you have until then to restrict who gets a dot and what it can do. After that, dots are on unless an admin opts out.

What changed on 4 November?

OpenAI's ChatGPT release notes for 4 November made three changes for organisations:

  1. General availability. Dots became generally available to eligible Enterprise, Edu and Healthcare workspaces.
  2. Admin preview until 18 November. Admins get a window to review and configure dots before the default flips.
  3. Admin controls. Workspace owners can now control who can use dots and which computer capabilities and actions are allowed, as shattered.io summarises from the release notes.

The same release let personal dots keep working on tasks between conversations.

This reverses the launch default. At the 29 September launch, CNBC reported that Enterprise access was an admin-enabled beta, off by default. Plenty of third-party guides still describe that older state, so check the date on anything you read.

  1. dots launch; Enterprise is an admin-enabled beta, off by default
  2. dots reach the EU on Business and Enterprise plans
  3. Enterprise, Edu and Healthcare GA; admin controls added
  4. Admin preview ends; dots on by default unless an admin opts out

Who is affected?

Eligible Enterprise, Edu and Healthcare workspaces. OpenAI's getting started article says dots aren't available for workspaces with UAE data residency or UAE inference residency, so those workspaces won't see the change.

Business Premium is a separate case. It includes one dot per eligible seat, and the 18 November default change is described for Enterprise, Edu and Healthcare. Business Standard seats don't include dots.

EU workspaces have had dots on Business and Enterprise plans since 28 October 2026. Consumer Pro dots still aren't offered in the UK, EEA or Switzerland, but that's a personal-plan rule and doesn't decide what an Enterprise workspace sees.

What should admins decide before 18 November?

  1. Decide who gets a dotChoose whether every member gets one, a pilot group does, or nobody does yet.
  2. Set computer capabilitiesDecide whether dots can use a local computer and browser control for your users.
  3. Restrict actionsChoose which actions dots may take, and which need a person to approve.
  4. Review connected appsPlugin permissions are shared across ChatGPT, ChatGPT Work and Codex, so audit them once.
  5. Tell your usersExplain what a dot can do, what it can't, and who to contact if one does something unexpected.

OpenAI hasn't published a click-by-click admin path that we could confirm, so the steps above describe the decisions rather than menu names. Look for the dots controls in your workspace admin settings. If your organisation wants dots off for everyone while it reviews them, make that change before the preview ends.

What controls do individual users have?

Admin settings set the outer limits. Within them, each person can tighten how their own dot behaves.

Control What it does Who sets it
Who can use dots Turns dots on or off for members Workspace owner
Computer capabilities Limits local computer and browser use Workspace owner
Allowed actions Restricts what dots may do Workspace owner
Custom Rules Act, act if pre-approved, ask first, or hand off, per action User
Local computer Off by default; user connects one or more machines User
Activity View Watch, redirect or stop a dot from the desktop app User

Some limits hold whatever the settings say. Password changes and money transfers always go back to the user, as eesel reports, and approving one action doesn't grant ongoing permission. When a dot does proactive research on its own, it is read-only: VentureBeat reports that during that work it can't send messages, change app content or control a browser. For a user-level walkthrough, see setting up dots Custom Rules.

Is workspace data used for training?

Not by default. OpenAI's dots privacy FAQ says Business, Enterprise and Edu data isn't used for training by default. OpenAI also says it doesn't train on a dot's proactive research or its private notes. Human review can still happen in limited cases, including safety cases.

Memory deserves attention in an organisation. A dot receives memories from ChatGPT and forms its own, including from connected apps. Individual dot memories can't be viewed, edited or deleted; the only way to clear them is to reset or delete the dot. Turning off ChatGPT Memory stops new sharing but doesn't remove what the dot already received. OpenAI hasn't published retention periods for dots.

In ChatGPT Space, anything copied onto a shared Page, including content drawn from memory, is visible to everyone with access to that Page. That's worth covering in any user guidance.

What about Microsoft Agent 365 and specialist dots?

OpenAI has said it is working with Microsoft to bring dots, especially specialist dots, under Agent 365 security and governance controls. As of November 2026 that has been announced but not shipped. Specialist dots get their own organisational identity, credentials and system access, unlike personal dots, which work under each user's own connected accounts. Specialist dot pricing hasn't been published.

Pros

  • Admins can limit who gets dots and which capabilities they use
  • Business and Enterprise data isn't used for training by default
  • Proactive research is read-only, and money transfers always go back to the user

Cons

  • Dots switch on by default after 18 November unless an admin opts out
  • Individual dot memories can't be edited or deleted
  • Agent 365 integration is announced but not shipped

Bottom line

If you administer an Enterprise, Edu or Healthcare workspace, 18 November 2026 is the date that matters. Before then, decide who gets a dot, what computer access it has and which actions it can take, and tell your users what to expect. For the wider workplace view, see personal AI agents at work and our guides to agent permissions and approval rules compared. For what a dot does day to day, read our dots review and the setup guide.