OpenAI Dots

What Dots is for

OpenAI describes Dots as always-on agents with their own cloud computer. They can keep working between conversations and use apps that you connect. The rollout began on September 29, 2026 for eligible Pro and Business Premium users; Enterprise access is a beta that an administrator must enable. Availability therefore depends on plan and market.

A bounded first task

A sensible first assignment is a reversible monitoring workflow: ask a dot to read one connected feedback channel and prepare a weekly draft report without sending it or changing source data. The official FAQ says proactive research may read permitted connected sources but cannot directly send messages or modify content through plugins. This keeps the output reviewable; it is an editorial workflow suggestion, not a measured result.

Permissions matter more than the prompt

The safety description says each dot works in an isolated cloud workspace, and you choose which apps or computers it may access.

Background proactive research is limited to read-only tools: it cannot directly send messages, change connected-app content or control a browser or computer. Any follow-up action goes through the normal action rules and checks.

Custom Rules can further restrict supported actions, but they cannot disable required safeguards. The product is not currently available to users under 18.

Some sensitive steps require approval every time or must be handed back to the user. Dots can still make mistakes.