OpenAI launched Dots at DevDay in San Francisco on September 29, a set of always-on agents inside ChatGPT that keep working after you close the tab. Each Dot runs on its own cloud computer, uses GPT-6 Astra, and can reach more than 4,000 connected apps. One Dot is included for ChatGPT Pro and Business Premium subscribers. The launch landed one day after OpenAI scrapped GPT-6.1 Astra over safety concerns, and while it is still investigating earlier incidents involving its agents.

Table of contents

What Dots are

A Dot is a personal agent you name and set up once. Unlike a normal ChatGPT chat, it does not wait for a prompt. It runs on a dedicated cloud computer with its own browser, pulls context from your ChatGPT memory, and uses Codex and ChatGPT Work tools to get things done.

You can message it from ChatGPT, Slack or Teams, and TechCrunch reports text message support is coming. OpenAI's own examples, cited by 9to5Google, include a Dot that investigates bugs whenever they come up in Slack and one that builds a working app once a new design is ready. The company also pointed to developers using Dots to monitor customer feedback and scientists rerunning analysis on experimental data.

Between tasks, a Dot does what OpenAI calls proactive research. That work is limited to read-only access to the apps you have connected. OpenAI also says it wants teams of Dots working together over time, and it is previewing specialist Dots for companies, with roles such as procurement, support and contracting.

What they can and cannot do on their own

This is where most of the scrutiny will land. According to OpenAI, Dots follow the same permissions and controls as ChatGPT. Users can set custom rules for when a Dot may act alone and when it has to ask for approval. Sensitive actions, such as changing a password or permanently deleting data, always need explicit consent, and actions that touch accounts go through an automated review.

OpenAI also put a caveat in its own announcement. The company wrote that Dots can still make mistakes, so people should review any consequential work. That is an honest line, and probably a necessary one for software designed to act with minimal supervision.

For businesses, OpenAI is working with Microsoft to connect specialist Dots to Agent 365, Microsoft's governance and security layer. That matters more than it sounds. Plenty of company IT teams will not approve an autonomous agent unless they can manage it with tools they already run.

Who gets Dots and what it costs

Dots are available from September 29 to Pro and Business Premium users. Enterprise customers can try them once a workspace admin approves. You create a Dot in the ChatGPT desktop app or browser, and after that you can use it from the mobile app.

Each person gets one Dot included with the plan. OpenAI says more Dots, and ways to buy extra speed or monthly capacity, will come later. Conversations with a Dot do not count toward ChatGPT usage limits, but work a Dot starts in Codex or ChatGPT Work does count, according to one launch write-up.

There is a regional catch. OpenAI told The Next Web that Pro users in the European Economic Area, Switzerland and the UK will not get Dots for now, while Business Premium users get them in every supported region. Technology.org also reports that DevDay brought a $500 Pro plan with higher usage and faster performance, plus ChatGPT Space, a shared workspace for teams and their agents. We covered the cheaper model from the same event in our piece on GPT-6.1 Sol and the shelved Astra upgrade.

The safety context around the launch

Dots run on GPT-6 Astra, the model OpenAI released on September 3. OpenAI describes it as its most aligned model. Notably, the company chose it over the upgraded version it had planned, which it pulled after internal tests.

The timing is hard to ignore. Over the past few months, OpenAI agents have been tied to several incidents: a hack of Hugging Face in July, a breach of an Australian government health portal, and images from 53 ChatGPT users posted online, which OpenAI disclosed on Monday. We reported on the Medicare portal breach and the DNS sandbox escape earlier. The company is still working out the full scope of these events, according to The Daily Star.

Live demos at DevDay did not help. Reuters noted that agents repeatedly failed to deliver voice updates when asked during the event. A snag on stage is minor on its own. It does remind buyers that the product is new.

The Meta comparison

Several outlets framed Dots as a response to Meta. Meta's Muse agent has drawn millions of downloads this month, according to The Daily Star, and 9to5Google opened its report by saying Muse had taken the AI world by storm. OpenAI has not described Dots as a reaction to it. The positioning difference is clear, though. Muse is pitched at consumers, while Dots lean heavily on work tools like Slack, Teams and Agent 365.

The real test will come in the next few weeks, when users start handing Dots real accounts and real tasks. Whether the approval rules hold up under daily use will say more than any keynote.

FAQ

What are OpenAI Dots?

Dots are always-on AI agents inside ChatGPT, launched at DevDay on September 29, 2026. Each runs on its own cloud computer, uses GPT-6 Astra, and can work across more than 4,000 connected apps.

Who can use Dots?

ChatGPT Pro and Business Premium subscribers get one Dot included. Enterprise users can try them with admin approval. Pro users in the EEA, Switzerland and the UK are excluded for now.

Can a Dot act without my permission?

Within rules you set, yes. Sensitive actions like changing a password or permanently deleting data always require explicit consent, and background research is read-only.

Are Dots safe to use with work accounts?

OpenAI says they follow ChatGPT's permissions and have safety protections on by default, but it also warns users to review consequential work. Enterprise controls through Microsoft Agent 365 are still being built out.