OpenAI is rolling out a new kind of AI agent designed to keep working even after a user leaves the conversation. Called Dots, the always-on agents can be given an ongoing job, work across connected apps and continue making progress between conversations instead of waiting for another prompt.
Each Dot is powered by OpenAI’s GPT-6 Astra model and gets its own cloud computer. Users give the agent a goal, choose which apps it can access and set rules for what it can do on its own. A Dot can work through multi-step tasks, decide what needs to happen next and return to the user when a decision requires human judgment.
That could make Dots useful for work that never really ends. An agent could keep up with a project, research changing information, handle recurring work or follow developments across connected services. OpenAI says Dots can use permitted plugins and can continue authorized work in the background. Users can also set boundaries, such as allowing an agent to read information from an app while preventing it from taking certain actions.
The added independence also creates new risks. OpenAI warns that websites, emails and documents can contain malicious instructions designed to trick an AI agent into revealing information or taking an unwanted action. The company says Dots use permission controls, automatic safety checks, approval requirements and monitoring to reduce those risks, but acknowledges that the protections cannot prevent every mistake.
Dots are beginning a gradual rollout to eligible ChatGPT Pro and Business Premium users, while Enterprise customers can access a beta if their organization enables it. OpenAI says the agents are initially available through ChatGPT on desktop web and the desktop app, with access varying by plan and market. The larger change may be what Dots represent: instead of AI being something users repeatedly stop and ask for help, OpenAI is building AI that can be assigned responsibility and keep working toward a goal.
