Key facts at a glance

• OpenAI has launched Dots, always-on AI assistants positioned as a direct response to Meta's Muse.
• Dots run on OpenAI's GPT-6 Astra model and use a dedicated cloud computer to browse the web and work across more than 4,000 supported apps.
• Users can interact through a text-message-like interface, voice calls via ChatGPT on web, desktop, or mobile, and integrations with Microsoft Teams and Slack.
• At launch, users can create only one Dot, though OpenAI plans to let people deploy multiple agents simultaneously.
• Dots provide progress updates, ask questions, learn preferences over time, and can suggest or complete tasks proactively.
• OpenAI is rolling out Dots to ChatGPT Pro, Business Premium, and Enterprise users starting today.
• The company is testing specialist Dots for organizations and says conversations with a Dot will not count toward ChatGPT usage limits.
• OpenAI says Dots include built-in and custom rules, plus an auto-review feature, to govern independent actions and permission requests.

OpenAI enters the always-on assistant race

OpenAI has launched Dots, a new class of always-on AI assistants that the company says can “do nearly anything” across connected apps while learning a user's preferences over time. The announcement, made during OpenAI's DevDay keynote, positions Dots as a direct response to Meta's Muse, a customizable AI assistant that has gained attention for browsing the web, making purchases, generating documents, tracking goals, and handling other everyday tasks. Like Muse, Dots are represented by cute, customizable avatars, but OpenAI is emphasizing a broader agentic role: the assistant is meant to work in the background, not simply answer prompts in a chat window.

The timing is notable. Meta's Muse arrived just weeks earlier, and its popularity has been accompanied by safety concerns. One user said Muse shared their address with a buyer on Facebook Marketplace without their knowledge. OpenAI is trying to get ahead of similar worries by pairing Dots with built-in rules, custom restrictions, and an auto-review process that checks actions before they are completed. The company's pitch is not just that Dots can do more, but that they can do more while remaining under user-defined control.

How Dots work

Dots are powered by OpenAI's GPT-6 Astra model. Each Dot has its own cloud computer, which it can use to access a web browser and more than 4,000 supported apps. That setup allows a Dot to move across services, gather context, and complete multi-step tasks without requiring the user to shuttle between tabs or manually copy information from one app to another. OpenAI describes the experience as an assistant that can operate in the background while the user focuses on other work.

Interaction is designed to feel familiar. Users can communicate with a Dot through a text-message-like interface, similar to the one offered by Muse. They can also start a voice call with the Dot from ChatGPT on the web, desktop, or mobile. Dots can connect to Microsoft Teams and Slack, where they carry over context from sessions on other apps and devices. OpenAI says the assistant will soon be able to communicate over text message as well, extending its reach beyond the ChatGPT interface.

The text-message-like format is important because it changes how people delegate work. Instead of opening a separate dashboard, users can send a request as if they were messaging a colleague. The Dot can reply with progress updates, ask clarifying questions, and present results in the same thread. That conversational layer is meant to make agentic AI feel less like a technical tool and more like a persistent working relationship.

What Dots can do

OpenAI offered several examples of how Dots might be used. A developer working on an app could ask their Dot to use customer feedback to build and test updates, then present a video showing the changes. A content creator could have their Dot review an interview transcript, identify moments worth clipping, create show notes, and write social posts. Those examples point to a wide range of knowledge-work tasks: research, content production, software testing, and coordination across apps.

Users can follow along with a Dot's progress by accessing its cloud computer. That visibility matters for trust. If an assistant is going to act independently, users need a way to see what it is doing, where it is doing it, and why. The cloud computer provides a window into the Dot's work, while the messaging interface provides a running commentary. Dots also learn while they work, allowing them to personalize output and come up with other tasks to complete “before you even think to ask,” according to OpenAI.

That proactive behavior is one of the most ambitious parts of the product. Most AI assistants today wait for a prompt. Dots are designed to anticipate needs, suggest next steps, and take on work that the user may not have explicitly requested. OpenAI says the assistant can learn preferences over time, which could make it more useful for recurring workflows. The same capability also raises questions about oversight, because an assistant that acts before being asked must be governed by clear rules.

One Dot for now, more later

At launch, users can create only one Dot. OpenAI plans to allow users to deploy multiple agents at once in the future. The company also says users will eventually be able to scale the output of each Dot by increasing its speed or the total amount of work it can take on per month. Those options suggest a future where individuals and teams manage a small fleet of specialized assistants rather than a single general-purpose bot.

OpenAI is already testing a feature that lets companies create “specialist” Dots to handle specific roles inside organizations. A specialist Dot might focus on a particular department, workflow, or set of responsibilities. That approach could help businesses deploy agents with narrower permissions and more targeted knowledge. It also fits the broader trend toward agentic AI in the workplace, where assistants are expected to take action across software systems rather than simply generate text.

Collaboration and context

Users and their colleagues can work with a Dot in ChatGPT Space, a new collaborative workspace inside the app. The workspace is designed to make the assistant part of a shared environment, not just a private chat. That could be useful for teams that want to hand off tasks, review a Dot's work, or coordinate around a project. Context from Teams and Slack adds another layer, allowing the Dot to carry information between communication tools and the tasks it is handling.

The emphasis on context is central to the product. A useful always-on assistant needs to understand what is happening across a user's apps and devices. It needs to know which project is active, which people are involved, and which information is relevant. Dots are designed to maintain that context as they move between services. OpenAI says the assistants can connect to more than 4,000 supported apps, a figure that signals broad integration ambitions.

Safety rules and auto-review

OpenAI says Dots come with built-in rules for when to act independently. Users can also set custom rules that block certain actions or dictate when the Dot should ask for permission. An auto-review feature is meant to verify whether an action complies with custom rules and safety requirements. The same feature helps Dots determine which tasks should be handed off to users, such as changing a password.

Those controls are likely to be closely scrutinized. Agentic assistants can be useful precisely because they take action, but that also means they can make mistakes with real consequences. A Dot with access to email, calendars, workplace chat, and customer feedback could do a lot of good, but it could also expose sensitive information or make unwanted changes. OpenAI's answer is a layered permission system: built-in guardrails, user-defined restrictions, and automated review before certain actions are completed.

The safety question is especially relevant given the concerns surrounding Meta's Muse. Muse has been celebrated for its range, but one incident involving a user's address being shared on Facebook Marketplace has raised alarms about how much control users have over these assistants. OpenAI is clearly trying to differentiate Dots by emphasizing rules and review. Whether those safeguards are sufficient will depend on how they work in practice and how easily users can configure them.

Availability and pricing

Starting today, OpenAI is rolling out Dots to ChatGPT Pro, Business Premium, and Enterprise users. The company says conversations with a Dot will not count toward ChatGPT usage limits. That detail could make Dots more attractive for users who plan to delegate long-running tasks or keep an assistant active throughout the day. The rollout begins with a limited set of plans, but the product's direction is clear: OpenAI wants Dots to become a persistent layer across work and personal apps.

The launch also fits OpenAI's broader DevDay messaging around agents and automation. The company has been pushing ChatGPT beyond simple question-and-answer interactions toward tools that can complete tasks, connect to external services, and operate with greater autonomy. Dots represent a more ambitious version of that vision, because they are always on and designed to work in the background. They are less like a chatbot you visit and more like an assistant that stays with you across apps.

Competitive pressure and open questions

Meta's Muse has helped set expectations for what consumers want from an AI assistant: a friendly interface, broad app access, and the ability to handle errands and projects. OpenAI's Dots borrow some of that formula, including the avatar-based presentation and the text-message-like interaction. But OpenAI is leaning harder into the idea of an assistant with its own cloud computer and a wide network of integrations. The competition is not just about which assistant can answer questions better; it is about which one can be trusted to act.

That trust will depend on transparency, reliability, and control. Users will want to know when a Dot is acting, what it is accessing, and how to stop it. Businesses will want administrative controls, audit trails, and clear boundaries around sensitive data. The success of Dots will therefore depend as much on governance as on model capability. OpenAI's auto-review feature and custom rules are a start, but the company will need to show that they work across complex, real-world workflows.

OpenAI says future updates will let users add more Dots and increase each Dot's speed or monthly work capacity. For now, the company is positioning Dots as always-on assistants that can do nearly anything across connected apps, learn preferences over time, and operate with user-defined guardrails. The launch marks a new phase in the race between OpenAI and Meta, and it makes clear that the next battleground for AI assistants is not the chat window but the background work that happens across the apps people use every day.

Source:The Verge News