OpenAI has quietly widened access to Dots, its always-on autonomous agents, by inviting a limited cohort of ChatGPT Plus subscribers into a test program. The move, confirmed on November 19, 2026 by Thibault Sottiaux, who leads engineering for Codex and ChatGPT, marks the first time the persistent agents have been offered outside the company’s most expensive tiers. Until now, Dots were confined to Pro, Business Premium and Enterprise plans, making this a notable shift in OpenAI’s packaging strategy.
- Expanded access: A small, undisclosed number of ChatGPT Plus users can now test Dots, while all eligible ChatGPT Business Standard members in the United States receive one personal dot each.
- Technical foundation: Each dot runs on its own dedicated cloud computer and is powered by OpenAI’s GPT-6 Astra and GPT-6.1 Terra models, connecting to more than 4,000 apps through plugins.
- Geographic limits: Dots remain unavailable in the European Economic Area, the UK and Switzerland, and OpenAI has not indicated when those markets will be included.
What Dots actually do
Dots are not ordinary chat sessions. Each one operates as a proactive agent with its own cloud-based workspace, allowing it to keep working on a task between conversations rather than waiting for a fresh prompt. According to OpenAI’s documentation, a dot can take a project from an initial idea to a usable output, drawing on ChatGPT memory to learn how a user prefers to work. Users can define what the agent is allowed to do autonomously, review its output, redirect it, or pause it at any moment.
The company positions the feature around practical, multi-step work: assembling a presentation from scattered data and customer feedback, turning an interview transcript into show notes and social posts, or scoping, writing and testing a code change. Because each dot has its own computer and access to connected apps, it can research, create and follow through without constant supervision. The agents also work across ChatGPT, Slack and Teams, a detail that points squarely at workplace coordination and client follow-ups.
“Heads up, we’re starting to test dots with a small number of ChatGPT Plus users,” Sottiaux wrote on X, adding that he has been “recently seeing a lot more positives around how good the proactivity has been after many improvements.”
A deliberate, tightly scoped rollout
The limited nature of the test appears intentional. OpenAI has not disclosed how many Plus subscribers are included, whether they receive the same usage allowances as Pro users, or when the experiment might widen. Access is decided account by account, and the public pricing page remains unchanged. For Business Standard, the release notes specify that eligible members in the United States each receive a single personal dot inside their workspace, subject to usage limits.
The same rollout brings a set of supporting features: iOS widgets, Shortcuts integration, iMessage and RCS communication, replies directly from iOS notifications, and custom instructions. OpenAI is also beta-testing a refreshed mobile interface designed to make working with dots smoother. The feature first appeared at OpenAI’s DevDay on September 29, 2026, meaning roughly seven weeks passed between the stage demo and the first real users gaining access.
The regional exclusion is the most consequential caveat. Dots are unavailable across the European Economic Area, the UK and Switzerland, a restriction that has applied since launch. OpenAI’s help center states plainly that support staff cannot manually enable the feature and that restarting the app will not grant access. For international users, that leaves a large share of potential customers waiting on the sidelines while the company gathers feedback on performance and safety.
Competitive pressure and strategic intent
The timing reflects a broader contest over persistent AI agents. Google has opened its Gemini Spark agent to paying subscribers in the United States, and Microsoft continues to push the Autopilot capabilities of its Copilot super-app. By lowering the entry point from the Pro tier to the $20-per-month Plus plan, OpenAI is widening the funnel and normalizing the idea that an assistant should act, not merely answer.
That strategy carries real risk. An agent that operates independently, reaches into thousands of applications and can control a user’s PC is as delicate as it is powerful. Errors or manipulation attempts would unfold without anyone watching. The European holdout is equally telling: the absence of the most autonomous agents in tightly regulated markets suggests OpenAI is still working out how the product fits with EU rules. For now, users in select markets are effectively serving as the test bed for technology that is advancing faster than the safeguards meant to accompany it.