OpenAI Dots work continuously — and need broad access
September 30, 2026
OpenAI Dots handle ongoing tasks on dedicated cloud computers and connect to more than 4,000 apps. Their usefulness grows alongside the risk of mistaken actions.
What this is about
OpenAI introduced Dots at DevDay on September 29, 2026: always-available AI agents that continue tasks in the background while the user is away. Each Dot runs on a dedicated cloud computer and can connect to more than 4,000 applications through plugins, according to OpenAI.
Dots initially launch for ChatGPT Pro, Business Premium, and Enterprise subscribers. OpenAI plans to bring them to consumers later. The idea goes beyond a new chatbot: a Dot is meant to remember preferences, preserve context across devices, and execute work steps independently.
What OpenAI Dots actually do
A user describes an ongoing goal, such as monitoring competitors, analyzing data, or continuing a software project. The Dot uses a browser, connected applications, and an isolated cloud computer. People communicate through a chat-like interface or voice calls; Slack and Microsoft Teams connections are designed to carry over context from other sessions.
OpenAI says the system is powered by GPT-6 Astra. The company presents Dots as always-available helpers that can perform recurring work and report back. Which actions can happen without renewed approval, however, depends on connected services, their permissions, and the relevant safety mechanisms.
Why it matters
Persistent agents move the boundary from advice to action. A conventional chatbot responds when someone asks a question. A background agent can meanwhile read files, operate websites, prepare messages, or execute code. That can save time, but it expands the consequences of a misunderstanding, a compromised connection, or a malicious webpage.
The potential reach is substantial: OpenAI said at DevDay that ChatGPT now has 1.2 billion weekly users. Dots are initially limited to expensive and business plans, but a later broad rollout could introduce millions of people to persistent agents. Permissions, logs, revocation, and human approvals therefore become core product features rather than side concerns.
In plain language
A Dot resembles an assistant who receives a desk, keys, and access to several filing cabinets. The assistant can keep working after you leave the office. That is useful when the instruction and keys are narrowly matched. An ambiguous assignment or an overly broad master key, however, can cause more damage than a wrong answer written on a note.
A practical example
A small company connects a Dot to its calendar, email, spreadsheets, and project board. The instruction is to sort 200 support messages every night and surface five urgent cases in the morning. The Dot categorizes 190 correctly but mislabels ten. If it only drafts suggestions, a person can catch the errors. If it may reply to customers and issue refunds directly, the same ten mistakes have immediate business effects. The key difference is therefore not only the model but the actions it is allowed to take.
Scope and limits
- OpenAI has not published independent long-term evidence showing Dots’ reliability and error rates across thousands of real workflows.
- More than 4,000 connections do not mean every integration offers identical safeguards, logs, or fine-grained permissions.
- Persistent agents are unsuitable for irreversible payments, publishing, or sensitive decisions without explicit approval gates and auditable logs.
SEO & GEO keywords
OpenAI Dots, AI agents, ChatGPT, GPT-6 Astra, background agent, cloud computer, app plugins, agent security, privacy, permissions, DevDay 2026
💡 In plain English
Dots are AI agents that keep working on cloud computers while the user is offline. Because they can operate many apps, their permissions need tighter control than those of a chatbot that only answers questions.
Key Takeaways
- →Each Dot works in the background on its own cloud computer.
- →OpenAI says more than 4,000 app connections are supported.
- →Dots initially launch for Pro, Business Premium, and Enterprise customers.
- →The agents are designed to preserve context across devices and connected services.
- →Broad access increases both usefulness and the potential cost of mistaken actions.
FAQ
Who can initially use OpenAI Dots?
Dots initially target ChatGPT Pro, Business Premium, and Enterprise subscribers. A later consumer release is planned.
Do Dots continue working while the user is offline?
Yes. Ongoing work on a dedicated cloud computer is central to the product.
Can Dots take actions independently?
They can operate connected applications and a browser. Their actual autonomy depends on permissions and approval rules.
Are Dots suitable for payments or publishing?
Only with narrow permissions, explicit approval gates, and auditable logs. Irreversible actions should not run unattended.