OpenAI 发布由 GPT-6 Astra 驱动的 Dots 智能体

OpenAI is rolling out Dots, always-on AI agents designed to own ongoing work instead of waiting for one prompt at a time. Powered by GPT-6 Astra, each dot has a cloud computer and browser and can connect to more than 4,000 apps through plugins. Users can name a primary dot, teach it through feedback, inspect its computer and assign multiple projects.

Dots can be messaged or called in ChatGPT on desktop, web and mobile, as well as reached through Slack and Teams, with texting planned. They carry context across channels and can send progress updates or questions. With permission, a dot can connect to other devices or work on a user's laptop.

A dot learns its owner's goals, preferences and standards, then returns completed work for review. OpenAI describes dots preparing tested software fixes, revising launch materials, updating scientific analysis, maintaining enterprise proposals and turning interview transcripts into content drafts. Inside OpenAI, they already investigate Slack bug reports, turn designs into apps and coordinate planning. One early tester's dot found a missed invoice and sent it after approval.

OpenAI is pairing that autonomy with controls for access and consequential actions. A dot's cloud computer stays separate from the user's machine unless access is granted, while saved passwords can be used on supported sites without being exposed to the model. Background proactive research is restricted to read-only tools, so it cannot send messages, alter app content or control a computer. Users can manage permissions, set Custom Rules, follow work in Activity View and redirect tasks. Auto-review checks actions against those rules and safety requirements, while steps such as changing a password remain with the user. Monitoring can pause or stop work over a safety concern.

The first dot is rolling out to ChatGPT Pro and Business Premium users in eligible markets at no extra cost, with an allowance for deeper work and extended limits during the first month. Enterprise users, including Edu and Healthcare workspaces, can enter the beta when an administrator enables it. Setup starts on desktop, with mobile access afterward. Conversations do not count against ChatGPT limits, although tasks run through Codex or ChatGPT Work use their usual allowances.

OpenAI is also previewing specialist dots for organizations, with dedicated identities, credentials and access to company systems. Focused enterprise pilots will build on internal testing in procurement, invoice processing, email marketing, support and contracting. A planned Microsoft Agent 365 integration would let businesses manage these agents through existing governance and security controls.

Source