OpenAI has started rolling out “dots,” a new kind of ChatGPT assistant that runs continuously on its own cloud computer and works on tasks in the background, according to the company’s launch post on its website. Each user gets one primary dot, which they can name, and which runs on GPT-6 Astra, OpenAI’s flagship model.
The pitch is a break from the chat window. A dot gets its own browser and its own set of connected apps, and the user can open that machine at any time to inspect what it did. OpenAI says dots can connect to more than 4,000 apps through its plugin ecosystem, and can take on several projects at once without the user managing separate threads. A dot takes messages and calls in the ChatGPT apps for computers, browsers and phones, and also answers in Slack and Teams. Texting is listed as coming later.
Availability is narrower than the headline suggests. Dots are live for Pro and Business Premium subscribers in eligible markets. Enterprise, Edu and Healthcare customers can try a beta once a workspace administrator switches it on. The first dot comes at no extra cost, and conversations with it do not count against ChatGPT usage limits. Tasks a dot hands to Codex or ChatGPT Work do count. OpenAI says users will later be able to buy additional dots or pay for more speed or volume, which points to where the pricing is headed.
The safety design is the part builders will scrutinize, and every description of it comes from OpenAI itself. The company says a dot’s cloud machine stays separate from the user’s own computer unless the user connects it, and that saved passwords can be used to sign in to websites without being shown to the model. When a dot works unprompted, which OpenAI calls “proactive research,” it is limited to read-only access to connected apps, so it cannot send messages or change content. An automated review step checks actions that could affect accounts or share information, and a monitoring system can pause or stop a dot. Changing a password is reserved for the human. OpenAI also concedes that dots “can still make mistakes” and tells users to review consequential work. The post links to a system card, but it offers no independent test results.
The evidence of usefulness is also self-reported. OpenAI describes staff using dots to start investigating bugs posted in Slack and to turn a design into a working app. It cites one outside early tester whose dot noticed an unsent invoice, drafted it, and sent it after his approval. That is a single anecdote, and the post gives no figures on accuracy, error rates or time saved.
The larger commercial bet is the second tier. “Specialist dots” would get their own identity, credentials and system access inside a company, and OpenAI says its internal testing covered five areas: customer support and commercial contracting first, then procurement, invoice handling and email marketing. These are an enterprise preview, not a product: OpenAI says it is starting with focused pilots in which its engineers work directly with each customer. The company is also working with Microsoft to bring specialist dots under Agent 365, Microsoft’s governance and security controls for agents. That matters because a software worker with its own login is a compliance question long before it is a productivity one.
For competitors, the move raises the bar from “an agent you invoke” to “an agent that is always on.” Anthropic, Google and the workflow vendors selling automation into Slack and Teams now face a rival that lives in the same channels and ships with a first seat included. The launch post gives no date for wider availability beyond “soon,” so companies weighing a pilot have no timeline to plan against, and should ask OpenAI for audit logs of a dot’s background activity before granting it access to anything that moves money.
Reported by OpenAI in its launch post “Introducing dots.” The post carries no publication date.
