OpenAI launches dots, an AI agent that works even while you sleep

OpenAI unveiled dots, an AI agent that operates continuously on its own cloud computer, during the DevDay 2026 conference in San Francisco on September 29 2026. The agent runs on GPT-6 Astra, the most powerful model available to the public, and is now accompanied by GPT-6.1 Sol. Each dot has its own cloud computer and browser, and can connect to more than 4 000 applications through plugins. Users can summon a dot from ChatGPT or from services such as Slack and Teams, and the agent can also interact with a user's personal computer when permitted.

Unlike traditional AI agents, dots act proactively, searching connected applications for ways to help without waiting for a request. OpenAI described an example in which a tester's dot noticed an unpaid publication, prepared an invoice, and sent it after approval. Access to dots is limited by subscription tier and geography; in France, a minimum subscription of 103 euros per month is required, with the first dot included at no extra cost for Pro and Business Premium plans. The service is available 24 hours a day, seven days a week, and includes a quota for deeper tasks that expands during the first month after launch.

OpenAI has published a security article outlining safeguards for autonomous agents, noting that disconnecting an application does not erase data already learned by a dot; only deletion removes that information. OpenAI also introduced specialized dots with distinct identities and access rights, and is collaborating with Microsoft to integrate these agents into the governance and security controls of Microsoft Agent 365.