OpenAI's 'dots' Are Here: Always-On AI Agents Inside ChatGPT

By Daniel Kong · October 2026

OpenAI's dots join Grok Bot, Meta's Muse and Microsoft's Copilot agents in the race to work on your behalf.

Within a month, every major AI platform has shipped the same promise: an agent that stops waiting for prompts and starts finishing work. OpenAI's entry — launched 29 September 2026 — is called dots.

The pitch is deliberately domestic. Instead of a tool you operate, a dot is a named colleague you message, and one that keeps working when you close the laptop.

What a Dot Actually Is

Each dot runs on its own cloud computer with its own browser, and connects to more than 4,000 apps through OpenAI's plugin ecosystem. It is powered by GPT-6 Astra, the agent-focused model OpenAI has been rolling out gradually rather than releasing to everyone at once.

• It lives where you already work. You can message or call your dot inside ChatGPT on desktop, web and mobile, and also in Slack and Teams, with texting to follow.
• It works on its own. A dot can carry a project forward while handling several others, and message you with progress, questions, or a decision it needs.
• It learns on the job. Dots take feedback and keep preferences, so the second run of a task needs less explanation than the first.
• You can watch it work. You can open your dot's computer at any time to inspect what it has been doing.

What It Costs in Ringgit

Dots are bundled, not sold separately. OpenAI's own documentation lists Business seats at the following rates, converted at USD1 = RM4.08:

Dots require the Pro or Business Premium tier, and your first dot is included at no extra cost, with an allowance for deeper work and extended limits in the first month after launch. Business seats carry a two-seat minimum. Conversations with your dot do not count against your ChatGPT usage limits, but tasks you hand to Codex or ChatGPT Work still do.

The first dot is set up on desktop, then messaged from the phone like any other colleague.

The Safety Design, and Why It Matters More Than the Demos

An always-on agent with its own machine and logins is a governance problem wearing a productivity badge. OpenAI's design answers four questions an SME owner should ask before switching it on.

What can it touch? Each dot works on its own cloud computer, and your computer and its files stay separate unless you deliberately connect them. For websites, dots can use saved passwords without exposing them to the model.

What can it do without asking? Custom Rules let you allow an action, require approval, or block it outright. Built-in safety requirements apply regardless of what you configure, and sensitive tasks such as changing a password always stay with you.

What is it doing in the background? So-called proactive research uses read-only access to the apps you have already connected, so it cannot send messages, change app content, or control your browser or computer on its own initiative. You can follow all of it in Activity View.

What if it goes wrong? Dots run an automatic review against your instructions and rules, and OpenAI says its monitoring can pause or stop a dot's work if it detects a safety concern. The company is explicit that dots can still make mistakes, so consequential work needs your review.

Approval gates are the practical difference between an assistant and an unsupervised one.

What This Means for a Malaysian SME

Strip away the launch language and dots are the same proposition as Grok Bot and Meta's Muse: hand over a repetitive multi-step job and get a finished outcome in the tool where the work lives. OpenAI's own example is instructive — an early tester's dot noticed an invoice to a publication had never been raised, prepared it, and sent it once the owner approved.

The discipline that makes dots useful is the same one that makes them safe. Start with one dot, one job, and approval switched on for anything outbound. Give it its own logins scoped to the systems it needs, keep the audit trail in Activity View, and leave payments and anything customer-facing under human sign-off until you have weeks of clean runs behind you.

There is also a data question. OpenAI says it does not use content from ChatGPT Business, Enterprise or Edu workspaces to improve its models by default, and personal plans have their own controls. Under Malaysia's PDPA, however, your business remains the data controller for anything a dot processes — the vendor's setting is not your compliance.

Finally, note where this is heading: OpenAI is previewing specialist dots with their own identity and credentials for enterprise pilots across procurement, invoice processing, email marketing, customer support and contracting, plus integration with Microsoft's Agent 365 governance controls. The agent is becoming a colleague with an employee record — and your policies should catch up before that happens.

Frequently Asked Questions