Table of Contents
ChatGPT's Dots Agent Just Went Default-On for Enterprise, Edu, and Healthcare
Somewhere in your ChatGPT workspace, a member can now spin up a standing agent that keeps acting after they close the laptop, and nobody on your admin team had to approve it. OpenAI ran a two-week admin preview for Dots in Enterprise, Edu, and Healthcare workspaces from November 4 to November 18, 2026. Once that window closed, Dots flipped to on by default for every eligible workspace that had not explicitly opted out, per OpenAI's enterprise release notes. If your org fits that description and nobody logged the decision in a change-management ticket, you are now running an always-on agent whether you planned for it or not.
We covered what Dots actually does when OpenAI launched it at DevDay in September, in our writeup on the original launch. This guide skips the product tour and goes straight to the part admins need: who is exposed right now, what it costs, the Admin Console path to shut it off, and the specific limits that change the calculus for a Healthcare workspace.
What flipped between DevDay and now
At DevDay, enterprise access to Dots was opt-in only. An admin had to turn it on before anyone in the workspace could use it, and it stayed off otherwise. The November 4 update reversed that polarity for Enterprise, Edu, and Healthcare: those tiers got general availability, wrapped in a preview period where the default was still off but every admin could see the toggle and decide, per the release notes. Fourteen days later, the default itself switched. An admin who did nothing during the preview now has Dots live, not pending.
That is a meaningfully different rollout pattern than the Slack or calendar integrations most IT teams are used to reviewing. Those typically ship off and stay off until someone requests them. Dots shipped off, announced a countdown, and then turned itself on.
Who gets it by default now
| Workspace tier | Default state after Nov 18 | UAE data or inference residency | HIPAA eligibility |
|---|---|---|---|
| Enterprise | On, unless opted out | Not available | N/A |
| Edu | On, unless opted out | Not available | N/A |
| Healthcare | On, unless opted out | Not available | HIPAA-eligible in covered workspaces, per OpenAI |
Source: OpenAI ChatGPT Enterprise and Edu release notes, November 4, 2026 entry.
The UAE exclusion is a hard gate, not a toggle. If your workspace runs UAE data or inference residency, Dots simply is not offered there yet, opt-in or not.
What it costs
OpenAI's release notes do not list a separate line-item price for Dots inside Enterprise, Edu, or Healthcare plans. Instead, usage ties into the workspace's existing daily-active-user count: a member only counts as a dot's active user on a given day if they send that dot at least one message, per OpenAI's dots-management docs. Background work a dot does on its own, between messages, does not move that meter. Where per-dot billing does apply, pausing or deleting a dot stops it, and restoring access resumes it, so the cost exposure tracks the toggle, not a flat per-seat add-on you'd need finance to pre-approve.
How to check your exposure and opt out
The opt-out lives in the same console path as every other workspace capability, which is easy to miss if you are expecting a dedicated "Dots" settings page.
- Open Admin Console > Permissions & roles.
- Select Permissions, then Workspace default.
- Open Workspace capabilities and find Use dots.
- Turn it off for the workspace default, then repeat the check for every custom role your org has defined.
- Save, and confirm the change against your audit log.
Step 4 is the part worth rereading. OpenAI's own documentation states that a role granting access can still let a member use Dots even if the workspace default has the permission turned off, per the dots-management docs. Flip the workspace default and stop there, and you have not actually opted out. You have just hidden the toggle from most of your users while anyone in a role with its own grant keeps full access. If your org runs more than a handful of custom roles, that is the gotcha that turns a five-minute setting change into an afternoon audit.
Removing a member's grant does not delete anything they have already done. It hides the dot from inventory, pauses it, and marks it disabled; the member cannot resume it until an admin restores the permission.
Before you flip it on for Healthcare specifically
Dots is HIPAA-eligible in covered workspaces, but eligibility covers OpenAI's handling of the dot itself, not every connected service a dot touches. Third-party apps and services a dot connects through, Slack, a calendar, a ticketing system, retain their own separate compliance coverage, per OpenAI. Enabling Dots does not retroactively make a connected tool covered if that tool's own agreement does not already say so.
Three other settings matter before a Healthcare admin signs off:
- Allow local computer access is off by default for Enterprise, letting a dot read a member's local files and run commands on their machine only when an admin explicitly turns it on. Leave it off unless a specific workflow needs it, and review it again before the next access audit, since it is not inherited from HIPAA status.
- Use custom rules for dots is also off by default. It lets members write rules that guide a dot's actions and confirmations, and those custom rules cannot override OpenAI's built-in safeguards regardless of how they're written.
- Agent Security has to already permit on-request approval with auto-review as a reviewer before the Use dots toggle does anything useful. Check Admin Console > Agent Security > Global > Requirements first; flipping on Dots against an incompatible policy just leaves the feature inert for your users, not blocked with an obvious error.
Data residency changed too: new dots created after general availability support data residency and Enterprise Key Management, with inference residency in the EU or US. Dots created before the GA date move to the required region within 30 days of launch, per the release notes. If your org has an EU residency requirement and users who created dots during the earlier beta, confirm that 30-day migration actually landed before you treat the residency box as checked.
If an always-on agent isn't the move yet
Opting out workspace-wide is the blunt option. A sharper one is leaving the workspace default off and granting access only through a scoped custom role for the specific team that asked for it, which keeps the rest of your org on the old opt-in behavior without you having to revisit the setting every time OpenAI ships a new default.
If the standing-agent model itself is the problem, not just the timing, two adjacent approaches cost less to walk back. Salesforce's Paige is an employee-facing agent billed under a flat per-user add-on instead of conversation-based usage, which we broke down in our Agentforce named-agents guide. And if your real question is managed versus self-hosted agent infrastructure rather than any one vendor's rollout, our comparison of OpenAI's Agents API against Anthropic's managed agents covers the tradeoff directly.
What to do this week if you haven't touched this since DevDay
Pull up Admin Console > Permissions & roles today and check what Use dots is actually set to, in the workspace default and in every custom role, not just the one you remember configuring in September. If you're fine with Dots running, your remaining work is the Healthcare-specific checklist above: local computer access, custom rules, and the Agent Security prerequisite. If you're not fine with it, the opt-out is five clicks, but only if you remember to check the roles, not just the default. Set a recurring 90-day reminder to recheck it. OpenAI changed the default once already this quarter; nothing here says they won't do it again before your next planning cycle.