Before You Connect a ChatGPT Dot to Slack, Email, or Drive: What It Can Access and How to Revoke It (2026)
Short answer: A ChatGPT dot only gets the apps you connect on its Plugins screen. Review each app’s permissions before you connect it. Then set Custom rules so sending, deleting, or posting asks you first. Password changes and money transfers always come back to you. Disconnecting an app stops new access. It does not erase what the dot already learned from that app.
OpenAI’s new dots don’t wait for you to type. They run around the clock on their own cloud computer, with their own browser, and they can reach more than 4,000 apps through plugins.
That’s the whole appeal. It’s also why the setup screen matters more than the demo.
If you have ChatGPT Pro or Business Premium, a dot is probably already in your account or about to be. Here’s what one can actually touch once you link Slack, Gmail, or Google Drive, and the exact controls for scoping and pulling that access.
Jump to a Section
- Who Gets a ChatGPT Dot, and When?
- What Can a Dot Do Once You Connect Slack, Email, or Drive?
- How Do You Make a Dot Ask Before It Acts?
- A Starter Rule Set for the Three Riskiest Apps
- How Do You Revoke a Dot’s Access?
- Is a Dot Different on a Business or Enterprise Workspace?
- Tips
- Warnings
- FAQ
- Sources
Who Gets a ChatGPT Dot, and When?
ChatGPT dots, announced at OpenAI DevDay on September 29, 2026, are rolling out to Pro and Business Premium users, with the first dot included at no extra cost, according to The Next Web’s DevDay report.
There are a few catches on availability:
- Pro users in the EEA, Switzerland, and the UK are excluded at launch. Business Premium gets it in all supported ChatGPT regions.
- Enterprise, Edu, and Healthcare get a beta that stays off until a workspace admin turns it on.
- Under-18 accounts can’t use dots.
- One dot per plan for now. OpenAI talks about teams of dots later, but that isn’t live.
Dots run on GPT-6 Astra. You can message or call yours in ChatGPT on desktop, web, and mobile, or message it in Slack and Teams. Texting is in a limited beta for US Pro users.
Usage is the part people misread. Chatting with a dot doesn’t count toward your ChatGPT limits. Tasks it starts in Codex or ChatGPT Work do count, except during the first month, when OpenAI isn’t counting dot usage against eligible Pro, Business, and Enterprise allowances.
What Can a Dot Do Once You Connect Slack, Email, or Drive?
A ChatGPT dot can read, write, and act inside every app you connect on its Plugins screen, within the limits of that app’s permissions and your Custom rules; apps you never connect stay out of reach.
That last part is the good news. Nothing is wired up by default. You create the dot in the ChatGPT desktop app or on desktop web, then connect apps one at a time from the Plugins screen. Each app shows its permissions before you approve it.
Once an app is connected, the dot works in two modes:
- Background research. This is read-only. The dot can look through your connected apps, but it can’t message anyone, change content through a plugin, or drive a browser while doing it.
- Acting on a task. When you ask it to do something, or when a rule allows it, the dot can post, send, edit, and file things in the connected app.
When I compared the docs-based write-ups from DataCamp, DEV Community, and Nerds Chalk, I found that they all describe the same line between those two modes. Reading is broad. Acting is where your rules kick in.
Think about what “read” means for a Gmail or Drive connection. It means the dot can see the whole mailbox or every file that connection exposes, not just the thread you mentioned.
Honestly, that’s the part to decide on first.
How Do You Make a Dot Ask Before It Acts?
To make a ChatGPT dot ask before acting, open Settings, then Personalization, then the Permissions section (on mobile, Customize, then Custom rules), describe the action, and set it to “Ask before taking action” or “Hand off to you.”
Every rule gets one of four behaviors:
- Take action without asking. The dot just does it.
- Take action if pre-approved. It acts only when you’ve already said yes to that kind of task.
- Ask before taking action. You get an approval prompt each time.
- Hand off to you. The dot stops and you finish the step yourself.
Some limits sit above your rules. Changing a password and transferring money always hand off to you, whatever you set. Permanently deleting data and installing software may ask for approval each time. An auto-review also checks the recipient and the message before actions that affect your accounts, to catch a wrong address or something you didn’t mean to share.
Here’s where people get tripped up: custom rules don’t grant access. A rule that says “send email without asking” does nothing if Gmail isn’t connected. And no rule can switch off the built-in safety checks or loosen the read-only limit on background research.
A Starter Rule Set for the Three Riskiest Apps
The table below is a conservative starting point for the three connections the DevDay coverage kept coming back to: Slack, email, and cloud storage. Loosen a rule only after you’ve watched the dot handle that action well a few times.
| Action | Suggested rule | Why | Can you change it? |
|---|---|---|---|
| Read Slack channels, email, Drive files | Background research (read-only) | Harmless on its own, but scope what you connect | Only by not connecting the app |
| Post in a Slack channel | Ask before taking action | A wrong channel is public to the whole team | Yes |
| Reply to yourself or a draft folder | Take action without asking | Nothing leaves your account | Yes |
| Send email to someone outside your company | Hand off to you | Can’t be unsent | Yes |
| Share or move a Drive file | Ask before taking action | Sharing settings travel with the file | Yes |
| Permanently delete data, install software | Approval may be required each time (built-in) | Default safety rule | Can be made stricter |
| Change a password, transfer money | Always handed off to you | Fixed by OpenAI | No |
Table: built-in limits compiled from OpenAI’s dots help documentation as summarized by DataCamp and DEV Community, October 2026. “Suggested rule” is our recommendation, not an OpenAI default.
Two of the seven rows are fixed by OpenAI. The other five are on you.
How Do You Revoke a Dot’s Access?
To revoke a ChatGPT dot’s access, disconnect the app from the dot’s Plugins screen, use “Revoke access” in the ChatGPT desktop app for local computer access, and pause or reset the dot from the ••• menu on its profile if you want it to stop entirely.
Each of those does something different, so pick the right one:
- Disconnect one app. Open the dot’s profile, go to Plugins (Customize, then Plugins on mobile), and disconnect it. New access stops.
- Revoke computer access. Local machine access is optional and starts off. If you turned it on, confirm “Revoke access” in the desktop app to stop the dot from reading files or working on that computer.
- Pause the dot. From the ••• menu on its profile. Good for a vacation or while you rethink the rules.
- Reset the dot. This deletes the dot, including its conversations, saved memories, and scheduled tasks.
This is the part most people miss.
Disconnecting an app doesn’t delete what the dot already pulled into its context from that app. And per Nerds Chalk’s reading of the docs, you currently can’t view, edit, or delete individual dot memories. If a dot saw something it shouldn’t have, reset is the only full cleanup.
Also check the other side. The connection usually shows up in the third-party app too, such as Slack’s or Google’s list of connected apps. Removing it there is a good backstop.
Is a Dot Different on a Business or Enterprise Workspace?
On ChatGPT Business, Enterprise, and Edu workspaces, content is not used to train OpenAI’s models by default, and on Enterprise, Edu, and Healthcare the dots beta stays off until a workspace admin enables it.
That shifts the first decision to IT. If you’re the admin, OpenAI has a separate help page on managing dots in ChatGPT workspaces. Decide which plugins are allowed before you flip the switch, not after people have already connected their mailboxes.
On a personal Pro plan, your existing ChatGPT data controls decide whether your conversations help train models. Look at that setting before you connect a work account to a personal dot. Better yet, don’t mix the two.
When I checked the launch coverage for what a dot does when something goes wrong, I noticed that OpenAI describes a monitoring system that can pause or stop a dot if it spots a safety concern. Useful, but it’s a backstop. It doesn’t replace sensible rules.
Tips
- Connect one app first. Watch a week of activity before you add the next.
- Create a dedicated Slack channel for your dot’s updates so its posts never land in a busy team channel by accident.
- If your email provider supports it, connect a secondary mailbox or a label-scoped account instead of your main inbox.
- Write rules in plain, specific language, like “posting in #general” instead of “posting messages.”
- Review the dot’s activity after its first month, when the free usage window ends and its tasks start counting against your allowance.
Warnings
- “Read-only” background research still means the dot can read everything a connection exposes. Scope comes from what you connect, not from the rule list.
- OpenAI’s own launch materials say dots can still make mistakes. Review anything consequential before it goes out.
- Resetting a dot is permanent. Export anything useful from its conversations first.
- Don’t connect accounts you share with family or clients without telling them. The dot reads their messages too.
FAQ
Can I use a ChatGPT dot on the Plus plan?
No. At launch, dots are included only with Pro and Business Premium, plus an admin-enabled beta on Enterprise, Edu, and Healthcare. Pro users in the EEA, Switzerland, and the UK are excluded for now.
Does disconnecting Gmail delete what my dot already read?
No. Disconnecting an app stops new access, but the dot keeps what it already built into its context. Resetting the dot is the only way to clear its conversations and saved memories.
Can a dot change my password or send money if I allow it?
No. OpenAI keeps password changes and money transfers as hand-off actions that you must complete yourself, and custom rules cannot override that.
Sources
- OpenAI Help Center — Getting started with your dot
- The Next Web — OpenAI launches dots, always-on AI agents with their own cloud computers
- DataCamp — OpenAI Dots: Always-On Agents in ChatGPT, Explained
🔗 Related guides
Fact-checked against OpenAI’s dots help documentation, The Next Web’s DevDay 2026 coverage, and DataCamp’s documentation summary as of October 3, 2026. Dots are a new product and plan availability, limits, and menu names may change.
