OpenAI launched Dots on Sept. 29, introducing AI agents designed to take responsibility for ongoing work between conversations. Powered by GPT-6 Astra, each has its own cloud computer and browser and can use apps the user connects, according to OpenAI’s announcement.
The company describes agents that can investigate bugs, prepare launch materials or update research as new information arrives. Its plugin ecosystem connects to more than 4,000 apps. Those are OpenAI’s descriptions of the product’s capabilities, rather than independent performance findings.
Access starts with paid plans
OpenAI’s setup guide lists a gradual rollout to Pro and Business Premium users in eligible markets. Enterprise customers, including Edu and Healthcare, can try the beta when their workspace administrator enables it.
The first Dot comes with Pro or Business Premium at no additional charge, including an allowance for deeper work. Setup requires the desktop app or a desktop browser. Mobile conversations become available after setup where supported; users cannot currently create a Dot on mobile. Connecting a personal computer is optional and starts turned off.
Background research has narrower permissions
OpenAI distinguishes work a user has assigned from the research Dots initiate to find ways to help. In its safety explanation, the company says that proactive research uses read-only tools: it cannot directly send messages, change connected apps or control a browser or desktop. Any action prompted by those findings must pass the usual rules and checks.
A separate system called Auto-review checks actions such as sending emails or changing files against the user’s instructions and safety requirements. Custom Rules can add boundaries, but cannot remove mandatory protections. OpenAI acknowledges that Dots can still make mistakes.
There is also a limit to undoing access. OpenAI’s privacy FAQ explains that disconnecting an app prevents new access without erasing information already retained. Users currently cannot inspect or remove individual Dot memories; deleting the Dot clears its own context. Files and conversations it created elsewhere remain separately stored, as do memories shared with ChatGPT, which have their own controls.
