What Are Dots, OpenAI’s New AI Agents? Know Here
What if an AI could keep working on a project after you closed the chat instead of waiting for your next prompt? What is the idea behind What are Dots, OpenAI’s new AI agents? always-on agents designed to take responsibility for ongoing digital work.
OpenAI introduced Dots on September 29, 2026, describing them as agents that can pursue user goals across applications with limited supervision. Unlike a standard ChatGPT conversation, a Dot can keep making progress between conversations, use connected tools, work from its own cloud computer, and return to the user when it needs human judgment. Dots are powered by GPT-6 Astra.
What Are Dots, OpenAI’s New AI Agents?
So, what are Dots, OpenAI’s new AI agents? They are always-on agents built into ChatGPT that can take responsibility for ongoing tasks rather than simply answering one prompt at a time.
A user gives a Dot a goal, connects the applications it needs, and defines what it can do independently. The Dot can then work through the task, determine what needs to happen next, and bring its results back for review. OpenAI says each Dot has its own cloud computer and browser, along with access to connected tools.
The important distinction is persistence. A conventional chatbot generally waits for another instruction. A Dot is designed to keep the context of an ongoing project and continue working when appropriate.
How Do OpenAI Dots Work?
Dots combine a language model with tools, memory, computer access, and controls that govern how independently the agent can act.
GPT-6 Astra provides the underlying model, while the Dot’s cloud computer gives it an environment in which it can perform supported computer-based tasks. OpenAI says Dots can also handle recurring work, follow up through connected channels, and delegate parts of a task to subagents.
That changes the nature of an AI request. Instead of asking, “How do I complete this task?” a user can give the agent a broader objective and let it work through the individual steps. The Dot can decide what needs to happen next and ask the user to step in when a decision requires human judgment.
What Can OpenAI Dots Do?
The practical uses depend on the applications and permissions connected to a Dot, but OpenAI says the agents can work across services such as ChatGPT, email, and Slack. They can also handle recurring tasks and maintain context across ongoing work.
For example, a user could give a Dot a continuing research assignment and ask it to monitor relevant information, organize findings, and return with updates. A business team could use one for recurring administrative work or information gathering across connected applications.
Dots can also proactively review connected information and form memories from it, according to OpenAI’s Help Center. Users can schedule recurring work and reminders, while the Dot can be paused when they do not want it to continue.
Why Does the Cloud Computer Matter?
One of the most important technical details behind Dots is their cloud computer.
Rather than simply explaining how a person could perform an action, the agent has a computer environment in which it can carry out supported tasks. OpenAI says each Dot has its own cloud computer and browser.
This gives the agent a workspace for browsing and computer-based tasks while separating that environment from the user’s ordinary local computer.
The distinction is useful because an agent needs somewhere to operate. Giving it access to a browser or cloud computer allows it to move from generating instructions to carrying out parts of a workflow.
OpenAI also provides controls around computer access and connected applications, particularly for organizational accounts. That becomes increasingly important as agents gain the ability to take actions rather than simply generate text.
What About Privacy And Control?
More autonomy also creates a greater need for clear permissions and safeguards.
OpenAI says Dots can use connected applications only within the access and permissions available to them. Users can define what a Dot can do independently, while the system can ask for human input when a decision requires judgment.
OpenAI has also published specific safety evaluations for Dots because their persistent and proactive behavior creates different risks from a conventional chat session. Its testing includes scenarios involving malicious instructions in emails, unauthorized data disclosure, and actions that cross the boundaries of one task into another.
The published evaluations found no severe data-exfiltration breaches in some of the tested scenarios, but they also identified moderate scope violations in persistent multi-task environments. OpenAI reported that the flagged rate in one evaluation increased from 8.6% to 19.7% when the number of intervening tasks increased from five to ten. These were controlled evaluations, not measurements of ordinary production use, so the results should not be treated as a prediction of how often such behavior will occur in real-world use.
Who Can Use OpenAI Dots Right Now?
Dots are rolling out gradually rather than being available to every ChatGPT user.
As of October 1, 2026, OpenAI says Dots are available to eligible Pro users in supported markets, excluding the European Economic Area, Switzerland, and the United Kingdom at launch. They are also available to Business Premium users across supported ChatGPT regions, while Enterprise users can access a beta when their workspace administrator enables it. Enterprise access is initially off by default.
Users create a Dot through the ChatGPT desktop app or desktop web. A Dot can then be accessed through the ChatGPT mobile app where mobile access is available, but OpenAI says users cannot currently create Dots on mobile and Dots are not supported on mobile web.
OpenAI is also temporarily excluding eligible Dots usage from the plan allowances of Pro, Business, and Enterprise users during the first month of rollout.
Why Do Dots Matter For AI?
The significance of Dots comes from the shift from answering prompts to managing objectives.
Most AI interactions still follow a simple pattern: the person asks, the model responds, and the person decides what happens next. Dots are designed to remove some of those intermediate steps by allowing an agent to keep working toward a goal.
That could be useful for research, administration, monitoring, communication, and other digital tasks that require several steps over time. Instead of repeatedly reconstructing the context, the user can give the Dot an ongoing responsibility and return when there is something to review.
The approach also raises a practical question that goes beyond model intelligence: how much autonomy should people give an AI system? OpenAI’s design places considerable emphasis on permissions, confirmation policies, and monitoring because an agent that can act independently needs different safeguards from a system that only generates an answer.
Conclusion
The clearest answer to What are Dots, OpenAI’s new AI agents? is that they are OpenAI’s move toward AI systems that can take responsibility for ongoing work instead of waiting for a new prompt after every step.Powered by GPT-6 Astra and equipped with a dedicated cloud computer, connected tools, and persistent context, Dots are designed to plan, act, continue working, and involve the user when human judgment is needed.