Each Dot gets its own cloud computer and browser, with approval gates for sensitive actions

The Verge's Emma Roth describes Dots as OpenAI's answer to Meta's Muse: customizable, always-on agents that OpenAI says can "do nearly anything" across connected apps while learning a user's preferences. Each Dot runs on GPT-6 Astra and has its own cloud computer and browser, letting it operate websites and software rather than only answer questions.

Users can reach their Dot through chat-style messaging, voice in ChatGPT, and Slack or Teams, where it carries context from conversations; SMS is coming. Users get one Dot at launch, with multiple Dots planned later. Rollout covers Pro, Business Premium and Enterprise customers, and Dot conversations do not count against ChatGPT usage limits.

OpenAI's examples include a developer loop in which a Dot turns user feedback into a pull request with a video walkthrough, and a creator workflow that produces clips, show notes and social media drafts. Specialist enterprise Dots are in testing. Both examples involve a Dot producing finished work that a person then reviews, rather than acting entirely on its own.

Custom rules and an auto-review step gate sensitive actions, such as changing passwords. Those safeguards matter because the agent category has already had trust problems: The Verge notes a report that Muse shared a user's address on Marketplace without the user's knowledge. Related coverage includes Dots' Slack and Teams launch details and the battle to be your personal AI agent.

An agent with its own computer, browser and app connections can be powerful and risky. The quality of approval gates, how clearly users can see what a Dot did and how easily they can undo actions will determine whether the product earns trust. Excluding Dot chats from usage limits suggests OpenAI wants heavy use, which raises the importance of those controls.

Competition is also shifting to businesses. Meta has launched an enterprise AI platform, and coverage before DevDay argued OpenAI's agents needed to catch up. Whether specialist Dots leave testing, how auto-review performs in practice and whether incidents emerge will shape the outcome. More in our agents coverage and Meta coverage.