OpenAI has introduced Dots, an AI agent designed to work on tasks over time rather than respond only within a single chat. The agents run on OpenAI’s GPT-6 Astra model, have access to their own cloud-based computer, and can use connected apps to research, draft material, build software, and follow projects as requirements change.
Each user starts with one primary dot that can be named and customized. OpenAI says the product is rolling out to ChatGPT Pro and Business Premium subscribers in eligible markets. Enterprise users, including Edu and Healthcare customers, can test a beta version when a workspace administrator enables it. The company plans to let users add more dots later.
The central idea is persistence. A dot is intended to remember a user’s goals, preferences, standards, and feedback. And similar to other AI agents, users can assign a broader project and return later to review progress, answer questions, or approve actions.
From background research to completed drafts
OpenAI presents Dots as an assistant for work that evolves over days or weeks. A developer’s dot could monitor customer feedback, identify recurring bugs, prepare fixes, test them, and submit pull requests for review. A marketing or product lead could ask a dot to revise launch materials after a product scope changes. For content creators, OpenAI suggests the agent could identify clips in an interview transcript, write show notes, and draft social posts.
Dots can work across ChatGPT on desktop, web, and mobile. They can also be contacted through Slack and Microsoft Teams, while text messaging support is planned. OpenAI says context carries between these channels, so a project started in ChatGPT can continue in a team conversation.
According to OpenAI, each dot has its own browser and cloud computer. Users can inspect that computer at any time. They may also allow a dot to connect to a laptop, although OpenAI says a user’s device and its contents remain separate by default.
Platformer’s Casey Newton reports that, in an early hands-on test, he used a dot to draft emails, research details for an insurance form, prepare meeting materials, and contact a speaking agent after granting permission. He estimates that the agent completed work that would otherwise have taken him roughly two hours. The report also stresses that using such agents requires considerable trust, particularly when they receive access to email, documents, or other sensitive services.
Approvals remain central
OpenAI says Dots include several safeguards intended to limit that risk. When the agent works in the background, it can conduct what OpenAI calls proactive research using connected apps with read-only permissions. In that mode, it cannot send messages, edit app content, or control a user’s browser or computer.
For actions that could affect accounts or disclose information, Dots use an automatic review process. The system checks proposed actions against user instructions, custom rules, and OpenAI’s safety requirements. Users can define which actions a dot may take independently, which require approval, and which are blocked. Some sensitive steps, such as changing a password, always require the user to act personally.
OpenAI also says that safety monitoring can pause or stop work when it detects concerning behavior. The company nevertheless advises users to review consequential work because Dots can make mistakes.
For organizations, OpenAI is previewing specialist dots with their own identities, credentials, and access permissions. These agents would handle narrowly defined responsibilities, such as procurement, invoice processing, customer support, or contract work. OpenAI says it is beginning with enterprise pilots and is working with Microsoft on governance and security integration through Microsoft Agent 365.
Dots do not count against standard ChatGPT conversation limits, OpenAI says. However, work initiated through Codex or ChatGPT Work continues to count against the applicable usage limits.
Sources
- Dots – OpenAI
- OpenAI announces Dots, always-on agents powered by GPT-6 Astra with its own cloud computer, available in ChatGPT on web, mobile, and desktop to Pro users – Bloomberg
- OpenAI announces ChatGPT Space, a collaborative workspace for users and ChatGPT and Dots, available on Pro, Business and Enterprise plans – Engadget
- Hands-on with Dots, OpenAI’s work-focused agentic product: highly capable and intuitive, with natural-feeling conversations represented as a chat inside ChatGPT – Platformer
Stay up to date
AI for content creation: the latest tools, tips and trends. Every two weeks in your inbox:
