What are OpenAI dots? How the AI agent works and how to use it safely

What are OpenAI Dots?

Dots are customizable persistent AI agents from OpenAI that run on the latest GPT-6 Astra model. Each of the dots has their own cloud computer, browser, and connected apps, making them capable of running long continuous tasks rather than just answering prompts. 

According to OpenAI, their ability to carry on tasks independently is what makes dots a game changer in how you work with AI. A regular artificial intelligence chatbot reacts to prompts while a dot owns a responsibility, acting as your personal assistant. It’s like an intern who instead of just answering questions when asked, does the delegated management tasks, while coming back to you to clarify details or get your approval. The dot tracks the numbers, follows up with colleagues, and shows up with a finished draft.

When announcing dots on OpenAI DevDay 2026 on September 29, 2026, Sam Altman, the CEO of OpenAI presented this technology as a difference-maker in everyday management. In addition, dots' ability to work in the background, marks another shift from generative AI, which produces answers on request, to agentic AI, which carries out multiple tasks while keeping tabs on everything the user deems important.

How do OpenAI dots work

Open AI dots work by running on a dedicated cloud computer and browser (separate for each dot), working even when your laptop or phone is off. The setup is similar to a virtual machine running in the cloud, dedicated to one agent. 

Dots connect to more than 4,000 apps through plugins, along with tools such as Codex. They can also use a local Mac or PC, but only with user’s permission (proactive background research is read-only). The decision making remains in the hands of the user — when needed, the dot will reach out to ask for permission to execute tasks (for example, postpone meeting, approve AI-generated presentation, or set up a call). When work is finished, the dot will bring the results back. Dots also learn user preferences and standards based on context and feedback over time, allowing them to better adapt to the owner’s needs and workflow.

OpenAI users can set up their dot through ChatGPT on desktop, web, and mobile, or through Slack and Microsoft Teams. Dots users can also speak to them through voice call with a texting option coming in the foreseeable future. 

While it may not seem like dots are that much different from regular LLM agents, the difference becomes clear upon closer comparison. 

OpenAI dots vs. regular ChatGPT

The main difference between dots and ChatGPT agents is that ChatGPT is request-driven, while Dots are responsibility-driven. Below you can see a clear distinction between the features of these two technologies:

In other words, dots are something that works continuously in the background, managing your everyday tasks that you might find too mundane or time-consuming to do yourself, while ChatGPT works as a consultant, whenever you want to ask or find something. It’s like having an assistant that takes care of your secondary tasks vs. having a colleague that answers questions and provides tips, but only when you ask them.

What can you use OpenAI dots for?

Dots can be immensely useful when dealing with work that can run unattended. It’s perfect for monitoring, planning, and tracking tasks. For example, dots can be helpful to experts such as:

• Developers. Dots can help monitor GitHub issues and user feedback, drafting fixes or summaries for review.
• Product managers. Dots can track launch tasks and follow up with owners.
• Researchers. Dots can read new papers or data and summarize the findings.
• Sales experts. Dots keep deal notes current and draft follow-up emails.
• Content creators. Dots can track a publishing calendar and prepare drafts.

OpenAI describes an early tester whose dot noticed he had forgotten to invoice a publication. The dot prepared the invoice and sent it only after the tester approved it. OpenAI also previewed specialist dots for organizations, which take on defined responsibilities such as procurement or customer support. The company is currently working with Microsoft to manage specialist dots through Agent 365 for enterprise governance and security. 

It’s also important to mention that dots are a poor fit for one-off questions. For simple questions and quick answers, a regular chat is faster. Dots are far more useful in ongoing work that is triggered by events and can be described clearly.

OpenAI dots pricing and availability

At the moment, dots aren't available on free OpenAI plans. The feature is included at no extra cost on OpenAI’s Pro plan tier, starting at $100 a month. As for usage:

• Conversations with your dot don't count toward your normal ChatGPT usage limits.
• Tasks your dot starts in Codex or ChatGPT Work do count toward those limits.
• OpenAI mentions an allowance for “deeper work” with extended limits in the first month, but it hasn't published exact numbers.

As of writing of this article, dots are rolling out to Pro and Business Premium users in eligible markets, available to users above 18. Enterprise, Edu, and Healthcare users can try a beta version when a workspace administrator enables it. OpenAI Dots are not offered in the EEA, Switzerland, or the UK. 

How to set up and use OpenAI dots

To set up OpenAI dot open the latest version of ChatGPT mobile app on iOS or Android, the desktop app, or desktop web. Here’s how to set up your dot:

1. 1. Check eligibility. Confirm that your plan and region support dots.
2. 2. Create your dot. Open ChatGPT on desktop or in a desktop browser and start a new dot. Follow the onboarding prompts to get started.
3. 3. Define the goal. Name your dot and describe its ongoing responsibility, its scope, and how you want results delivered.
4. 4. Connect only what it needs. Link only the apps the goal requires, and set each one to read-only or action permissions.
5. 5. Set custom rules. Decide which actions are allowed, which need approval, and which are blocked.
6. 6. Choose channels. Pick ChatGPT, Slack, or Teams, and decide whether to enable voice.
7. 7. Review regularly. Check the Activity View and give your dot feedback.

If your dot isn't responding, check OpenAI's service status before changing your settings.

Are OpenAI dots safe?

According to OpenAI, the company has built meaningful safeguards into dots, and the product is designed so the user stays in control. As with any agent that can act on your behalf, though, how you set it up matters. The same logic applies to pros and cons of AI in general: While the benefits can be massive, some risks will always be there, even with the most privacy-focused approach.

Built-in safeguards

Based on OpenAI, the company implements these safeguards in regards to dot safety:

• Each dot runs on its own cloud computer, and the user's computer stays separate unless connected to the cloud.
• Dots can use saved passwords to sign in to supported websites without exposing them to the model.
• Proactive background research uses read-only tools by default, so it can't send messages or change app content.
• Custom rules let users manage permissions for specific actions.
• Activity View shows what the user's dot is doing, allowing them to redirect it to other tasks if needed.
• An auto-review system checks actions that could affect user’s accounts before executing them.
• Sensitive tasks, such as changing a password, are left to the user's discretion.

Custom rules work like this:

Potential risks to keep in mind

Even with all the safeguards, user should still keep these risks in mind:

• Mistakes. Agents and dots can still get things wrong. Users should still be vigilant when interacting with them. 
• Connected app risk. The more apps and permissions a dot has, the larger the impact of an error. It might be best to keep the connected app number to a minimum for a while.
• Prompt injection. A dot that reads emails, web pages, or documents could encounter hidden instructions meant to manipulate AI agents. Although OpenAI says its safeguards defend against malicious instructions and monitor for harmful behavior, no system is ever fully safe from threats. 
• Data handling. Dot's work and connected data are processed in OpenAI's cloud. However, the company claims it doesn't use content from ChatGPT Business, Enterprise, or Edu workspaces for AI data training by default, allowing users to control whether dot's conversations are used for model training. Make sure your data controls are set before you connect sensitive accounts. 
• Scams riding the hype. Fake dot apps, browser extensions, or phishing emails may increase in number following the hype this feature has caused. Attackers may also use AI tools themselves (dark AI), so users should be careful not to fall into the trap.

Tips for using OpenAI dots securely

While OpenAI is gradually rolling out dots to their Pro users, it’s good to remember some tips that may be useful when dealing with this feature:

1. 1. Grant the least privilege needed. Connect only the apps the goal requires, and keep them read-only where possible.
2. 2. Set up mandatory approval rules for risky actions. Anything that sends messages, spends money, shares files, or contacts people outside your organization should require your approval.
3. 3. Keep local access off by default. Enable Mac or PC access only when you need it, and limit it to specific folders.
4. 4. Keep sensitive data out of it. Don't share card numbers, government IDs, health records, or passwords unless the task truly requires them. This also answers the common question of what not to ask an AI.
5. 5. Protect your account. A dot is only as safe as the ChatGPT account controlling it, so use a strong, unique password and multi-factor authentication.
6. 6. Regularly review activity and connections. Check the Activity View and connected apps regularly, and remove plugins you no longer use.
7. 7. Follow work policies. At work, follow your company's AI policy and don't connect work tools to a personal dot. Unapproved AI agents are a form of shadow IT.
8. 8. Use official links only. Access dots only through official OpenAI apps and links, and be wary of emails or ads offering “early access.”

For an additional layer of defense, consider investing in services like NordVPN. For mobile users, its real-time protection blocks malicious domains, including phishing pages that imitate OpenAI. Meanwhile for desktop users, NordVPN's next-gen antivirus also scans files as you download them and blocks malware, taking a different approach from signature-based scanning. Along with a VPN, NordVPN looks for suspicious behavior, not only known risks, so it can catch new threats and protect your online presence.

Can you use OpenAI dots while traveling in the EU or UK?

At the moment, dots aren't offered in the EEA, Switzerland, or the UK, meaning it’s not possible to use dots while in these locations. However, because a dot runs on its own cloud computer, it should work while traveling, meaning users who have access to dots should be able to use them on their trips. 

However, even though dot availability is account-based (rather than based on IP address), the users might need to be in the US to access the dot. OpenAI checks your device's live IP address and actively hides the feature in EEA, Switzerland, and UK, with exception being Business Premium and Enterprise plan users, which is where VPN comes in handy. 

VPNs have servers in many countries, including the US, encrypting traffic between your device and the VPN server and replacing your IP address with the server's. Services like NordVPN also come with next-gen antivirus capable of blocking phishing links and malicious downloads. It’s a useful tool to safeguard yourself from snoopers when browsing public Wi-Fi, browsing online, or managing your dot. 

A VPN is not a way to get dots in a region where OpenAI doesn't offer them. Connecting from an excluded location to get around availability limits may conflict with OpenAI's terms of use, so read those terms before you travel.

What to keep in mind when using a VPN with dots

When using VPN with dots, it’s important to remember that:

• OpenAI's terms of use apply wherever you are, so review the availability terms before a trip.
• Signing in from a new location or IP address may trigger extra verification. This is a normal account-security check.
• A VPN protects data in transit between your device and the VPN server. It doesn't change how OpenAI stores or processes the data your dot handles.
• If a connection feels unstable, trying a different server in the same country might solve the issue.

OpenAI dots alternatives

Several companies now offer agents that work across apps and tasks. Each takes a different approach to where the agent runs and how much control you keep.

Below you’ll find a more detailed comparison of all available dot alternatives.

OpenAI dots vs. Meta Muse AI agent

Compared to OpenAI dots, the main difference between dots and Muse is focus. Muse (Meta’s personal AI agent announced September 8, 2026) is built for personal life and includes shopping, with one-time card numbers generated at checkout. Dots are built around work and ongoing responsibilities, and they fit into team tools such as Slack and Teams. Muse is available to users 18 and over in the US. Dots are available to eligible paid subscribers. 

OpenAI dots vs. Claude Duet

Similar to Muse, Claude Duet is Anthropic's proactive agent, released as a research preview on October 21, 2026. Similar to dots, Claude Duet also comes with approvals and is designed to ask before it sends a message, makes a payment, signs a document, or deletes anything. However, dots are more widely available since Claude Duet is access-limited due to regional rollouts, tier structures, and waitlist rules.

OpenClaw

OpenClaw is an open-source AI agent that you run on your own hardware. Compared to OpenAI dots, It may give users more control over their data, but it also requires more technical setup, manual security management. It works great for those willing to tinker with their security settings, but might not be a favorable option to an everyday user. 

Is OpenAI dots worth it?

Overall, dots seem like a logical next step of AI improvement. It has a potential to reduce the amount of mundane tasks (such as monitoring, follow-ups, and weekly reports), which is why it makes sense to have. However, at the same time, it also carries data privacy risks, since it requires a lot of personal data (calendar, contacts, business-sensitive information) to be able to function properly. Some might also doubt how much of an impact this feature will actually bring to everyday life. 

Ultimately, the decision is yours to make. To make it more simple, you can ask yourself three questions:

1. 1. What work needs ongoing memory?
2. 2. What could safely run while you're away?
3. 3. Can you describe the outcome clearly?

If your answers point to work that runs in the background, a dot may be worth trying. However, if you only use AI to ask questions, a simple LLM might save you more time than a constantly awake virtual assistant.

Whichever option you choose, make sure to protect the connection and accounts behind it. Try NordVPN for more privacy and security online.