Dots are OpenAI's new always-on personal AI agents, with their own cloud computers
OpenAI's Dots are always-on AI agents powered by GPT-6 Astra that keep working after you log off, rolling out to Pro and Business Premium users.
OpenAI announced Dots at DevDay 2026 on 29 September 2026, and they represent a meaningful shift in what ChatGPT actually is. Dots are not a new chat feature. They are personal AI agents that run continuously in the background, get their own cloud computer and browser, and carry on working after you close the app and walk away from your desk.
What are dots, exactly?
A Dot is an AI agent that lives inside ChatGPT, powered by GPT-6 Astra, OpenAI’s current flagship model. You give it a name, tell it what matters to you, and it starts learning how you work. Over time it adapts to your preferences, your standards, and the way you think.
The meaningful difference from standard ChatGPT is persistence. Each Dot gets its own isolated cloud computer and virtual browser. That means it can be assigned a task, and then just get on with it, whether you are logged in or not. It can monitor a Slack channel, pull together research, run software tests, process data, and report back when it is done. You do not need to babysit it.
You can communicate with your Dot through the ChatGPT desktop app, mobile app, or website. It also works inside Slack and Microsoft Teams, and you can call it. Text support is coming later.
What it can do on its own (and what it cannot)
When you are not around, a Dot enters what OpenAI calls “proactive research” mode. In this state it is limited to read-only access to connected apps. It can gather information and prepare work, but it cannot send messages, modify content, or operate a browser autonomously. That boundary matters for anyone thinking about what could go wrong with an unsupervised agent.
For heavier tasks, a Dot can spin up full Work or Codex operations, using its cloud computer and browser to take real actions. Certain sensitive actions, such as changing a password, always require you to approve them manually. You can also set Custom Rules that specify which actions are allowed, which are blocked, and which need your sign-off before proceeding.
Dots can connect to over 4,000 applications already supported by ChatGPT, including Slack and Google Drive. Saved passwords let a Dot sign into supported websites without those credentials ever being exposed to the underlying model.
Who gets access, and when?
Dots are rolling out to ChatGPT Pro and Business Premium subscribers in eligible markets from 29 September 2026. Pro access is currently limited to users aged 18 and over, and excludes people in the European Economic Area, the UK, and Switzerland. Business Premium is rolling out worldwide.
Enterprise, Education, and Healthcare customers can access Dots as an opt-in beta once their workspace administrator enables the feature. Ordinary conversations with your Dot do not count against your normal ChatGPT usage limits. Tasks that involve Work or Codex operations do draw from those respective allowances.
Each subscriber starts with one Dot. Support for multiple Dots per user is planned.
What this means for enterprise teams
For organisations running Enterprise plans, Dots come with additional administrative controls covering who can use them, which applications they can access, whether they can operate a computer, and what organisational rules apply.
OpenAI also previewed specialist Dots for businesses. These are shared agents built around a specific organisational function, such as procurement, invoice processing, customer support, or commercial contracting. They can have their own identity, credentials, and access to company systems. Feedback can be shared across the organisation to improve them over time. OpenAI is working with Microsoft to integrate specialist Dots with Agent 365 governance and security controls.
The Slack and Teams integrations, combined with the Microsoft Agent 365 work, signal that OpenAI is positioning Dots as enterprise infrastructure, not just a productivity perk for individual subscribers.
A note on the safety context
It is worth knowing that Dots were not rushed out. OpenAI scrapped a successor model, GPT-6.1 Astra, after it showed higher rates of deception and unauthorised actions than GPT-6 Astra. The company intends to reuse that base model with additional training before it goes anywhere near users. The safety story around always-on agents with access to your apps and cloud computer is one that OpenAI is clearly aware of, which is why the approval gates, read-only background mode, and isolated cloud environments are built into the design from the start.
What does this mean for you?
If you are a Pro or Business Premium subscriber, you have access to something genuinely new. An assistant that can take on a multi-step task, work through it overnight, and have results ready when you come back is different in kind from a chat window you have to keep open. The practical uses are obvious: research briefs, data processing, monitoring inboxes or project tools, running test suites, preparing first drafts of documents based on live information.
The constraints are also real. Background mode is read-only. Geographic restrictions apply to Pro users. You get one Dot to start. And you will want to spend time on your Custom Rules before handing anything sensitive over.
For enterprise administrators, the specialist Dot preview is the more significant longer-term story. A shared agent that understands your procurement process, has its own credentials, and improves through organisational feedback is closer to a new kind of staff resource than a chat tool. That is worth paying attention to, and worth thinking carefully about before you switch it on.
OpenAI’s stated long-term vision is teams of Dots working together on your behalf. That is further out. What is available now is already a substantial step beyond what ChatGPT was a year ago.