OpenAI Launches Dots Always-On AI Agents
OpenAI released Dots, cloud-based AI agents powered by GPT-6 Astra that perform tasks and manage workflows independently of user devices.
OpenAI launched Dots on September 29 during its annual Developer Day. Powered by the GPT-6 Astra model, Dots are always-on AI agents that operate on independent cloud computers, enabling them to investigate problems and manage workflows even when a user's device is offline. The product is available to Enterprise, Business Premium, and Pro subscribers, though the initial Pro rollout excludes Switzerland, the United Kingdom, and the European Economic Area.
The agents evolve from the open-source OpenClaw project and compete directly with Meta's Muse personal agent. While positioned as enterprise tools for project management and bug monitoring, security analysts have noted that governance of the agents remains primarily with the vendor rather than internal corporate security teams.
To manage safety, the company implemented custom rules and auto-review systems to limit agent autonomy. OpenAI acknowledged the agents are not infallible, reporting that they passed 45 of 49 tests regarding permission changes. CEO Sam Altman described the safety approach as an "abundance-of-caution thing," noting that the model performed slightly worse on a few specific evaluations.
