OpenAI's Dots take on work tasks from their own cloud computers

OpenAI unveiled Dots at its DevDay 2026 conference, positioning them as always-on agents that operate from their own cloud computers. Unlike the existing Computer Use feature in ChatGPT Work, where users initiate each task individually, Dots run on separate machines and can take over ongoing work. They come with a browser, terminal, and file storage, and can connect to over 4,000 apps via plugins. OpenAI says Dots learn from feedback how their user works and what they consider a good result.

The launch includes a cheaper model, GPT-6.1 Sol, while the top-end GPT-6.1 Astra is being held back due to safety problems. Dots themselves run on GPT-6 Astra. In terms of safety, Dots operate under a set of ground rules that govern when they act alone and when they need approval. Background "proactive research" uses only read-only tools, and sensitive actions like password changes always stay with the human. OpenAI acknowledges Dots can make mistakes and recommends double-checking results with real consequences.

Access varies by region and plan. Pro subscribers in Europe must wait, as Dots are limited to users outside the European Economic Area, Switzerland, and the UK for now. Business Premium customers get them in all supported regions, while Enterprise, Edu, and Healthcare customers can try a beta once admins enable it. Each user gets one Dot at launch, but OpenAI's long-term plan is whole teams of Dots working together. The company is also working with Microsoft to bring specialized Dots to Agent 365, its management platform, starting with pilot projects at individual companies.