OpenAI’s dots are always-on AI agents inside ChatGPT. Each one runs on its own cloud computer with its own browser, so it can keep working toward a goal after you close the app. According to OpenAI’s launch announcement, they are powered by GPT-6 Astra and can connect to more than 4,000 apps through the company’s plugin system.
In plain terms, a dot is less like a chatbot you consult and more like an assistant you assign work to. That difference raises practical questions about what it can do on its own, what needs your approval, and who can use it right now. The answers below come from OpenAI’s announcement, its safety post and its help pages, as relayed by Unite.AI.
What a dot is and where it runs
OpenAI announced dots on September 29, 2026, at its DevDay event, and the rollout is gradual. Each dot works on its own cloud computer, alongside the apps you connect, and you can open that computer at any time to inspect what the dot has done. Users start with a primary dot that they can name, and OpenAI says it expects teams of dots to work together on a user’s behalf later on.
A dot can also use your own laptop if you allow it. According to OpenAI’s getting-started guide, local computer access starts turned off and is enabled through the ChatGPT desktop app with an Allow access confirmation.
You can reach a dot in ChatGPT on desktop, web and mobile, and in Slack and Teams. OpenAI says it carries context across those channels. Texting is a limited beta for Pro users in the U.S., and dots cannot start calls to you at launch.
What OpenAI says dots can do
OpenAI’s examples are practical rather than dramatic. A dot can turn recurring customer feedback into tested code changes, revise launch materials when a project’s scope shifts, rerun a scientist’s analyses as new data arrives, update sales proposals, or prepare clips and show notes from interview transcripts.
One story from an early tester stands out. His dot noticed he had forgotten to invoice a publication, prepared the invoice and sent it after he approved it. A dot can also stay on one project while working on several others at the same time.
These are OpenAI’s own examples, not independent results. They show the intended use, which is steady background work with a person approving the important steps.
How approvals and safeguards work
OpenAI describes its controls in a safety, security and privacy post. When you are not working with a dot, it can look for ways to help through what OpenAI calls proactive research. Those tasks use read-only tools restricted in code, so they cannot send messages, change content in connected apps or control a browser or computer.
Before a dot sends an email or changes a file, a separate system called Auto-review checks the planned steps against your instructions, your Custom Rules and OpenAI’s safety requirements. If a step is blocked, the dot receives the reason and can ask for approval, try an allowed alternative or stop.
Custom Rules let you allow specific actions, require approval or block them, but some safeguards cannot be removed. Changing a password or moving money between financial accounts must be handed back to you. Permanently deleting data or installing unrecognized software needs confirmation each time, and purchases with cards saved on merchant websites need approval.
The risks OpenAI itself acknowledges
OpenAI’s privacy, security and safety FAQ says protections against prompt injection reduce the risk but do not eliminate it. Prompt injection means hidden instructions in a web page or document that try to trick an AI into taking an action its owner never intended. The same FAQ says dots can still make mistakes.
Sign-in is handled separately. OpenAI says credentials entered through its secure sign-in go directly to the browser environment, while a secret pasted into a message or document may still be visible to the model.
Privacy details worth knowing
A dot receives memories and recent conversation context from ChatGPT, and conversations with a dot can feed ChatGPT memory. Turning off the Memory setting stops that sharing but does not delete what the dot already has. A dot’s context can be deleted only by deleting the dot, and individual dot memories cannot currently be viewed or edited. Dots are not yet available to users under 18.
On training, OpenAI says it does not use content from ChatGPT Business, Enterprise or Edu workspaces by default. On personal plans, the “Improve the model for everyone” setting controls whether a dot’s work may be used.
Who can use dots today, and what they cost
Dots began rolling out to Pro users in markets that exclude the European Economic Area, Switzerland and the United Kingdom, and to Business Premium users across supported regions. Enterprise customers, including Edu and Healthcare, can try a beta once a workspace admin turns it on, and it is off by default.
The first dot is included in Pro and Business Premium at no extra cost. For the first month after launch, dots usage will not count toward plan allowances, and OpenAI says it will share usage terms afterward. Conversations with a dot do not count toward ChatGPT limits, but tasks it starts in Codex or ChatGPT Work do. Dots are created in the desktop app or on desktop web, not on mobile.
What this means for you
If you are on Pro or Business Premium in an eligible market, start with low-stakes, repeating work such as research, drafts and summaries. Leave local computer access off until you have a reason to turn it on, and use Custom Rules to require approval for anything you would want to review. Open the dot’s cloud computer now and then to see what it did.
If you are on a free or Plus plan, or in the EEA, Switzerland or the U.K., dots are not available to you yet.
A dot is not for people who simply want a quick answer, or for anyone uncomfortable giving software access to accounts and saved passwords. Since a dot’s context can only be removed by deleting it, think of that as part of the setup decision.
The practical question is not whether a dot can do a task, but whether you are comfortable with the approvals you set. Start small, read what it does, and widen its access only as that review gives you reason to.
