OpenAI's launch post describes an early tester whose dot noticed he had forgotten to invoice a publication. The dot prepared the invoice, and it went out after he approved.
Notice what the dot did and did not do. It found the gap and drafted the fix, and a person pressed send. That is what OpenAI dots are: always-on AI agents inside ChatGPT that keep making progress between conversations and message you when a decision needs you.
OpenAI launched dots at DevDay on 29 September 2026. As of 8 October 2026, Pro users and Business Premium seats in Malaysia and Singapore are eligible, though rollout is gradual.
The harder question is what a dot should be allowed to touch.
What a dot is, in plain English
A dot is your own agent in ChatGPT. You point it at a goal and connect the apps it may use. Then you decide which actions it takes alone and which need your approval.
What sets it apart from a normal chat:
- It has its own computer. Each dot runs on its own cloud computer with its own browser. You can open that computer at any time to inspect its work.
- It keeps going. It can take on ongoing responsibility, including scheduled jobs such as reviewing your calendar each morning.
- It reaches your apps. Through ChatGPT plugins it can connect to over 4,000 apps.
OpenAI says dots are powered by GPT-6 Astra, the model it introduced on 3 September 2026 (ChatGPT release notes). Your own computer is separate by default: access to it is optional and starts turned off.
What a dot can do: three illustrations
These are our illustrations of how a business here might use a dot, not OpenAI case studies. The right set-up depends on which apps you connect.
1. A sales manager's morning brief (Shah Alam distributor). She asks her dot to review her calendar each morning, along with her connected inbox. It lists today's customer visits and flags a RM48,000 quotation that has gone a week without a reply. Then it drafts a follow-up. The drafting is the dot's work; the sending is her decision.
2. Chasing unpaid invoices (Penang services firm). If the firm's accounting app has a ChatGPT plugin, the finance executive can have the dot check it each Monday for invoices past 30 days. It prepares reminders with the amounts in RM and holds them for approval, much like OpenAI's own invoice example.
3. Watching competitor prices (Singapore online seller). A seller on Shopee and Lazada asks the dot to check ten competitor listings every Thursday night, before the weekend sale, and summarise price changes in SGD. It reads the public pages with its own browser. Changing her own listings stays her job.
In all three, the dot does the slow gathering and drafting, while a person keeps every action that touches customers or money.
ChatGPT dots: which plans include them in Malaysia and Singapore
Short version: Pro and Business Premium users in Malaysia and Singapore are eligible, with rollout still gradual.
At launch, two paid plans include dots:
- ChatGPT Pro, in every market except the European Economic Area, Switzerland and the UK.
- ChatGPT Business with a Premium seat, in all supported ChatGPT regions. OpenAI's business pricing page lists dots against Business Premium specifically.
Enterprise, Edu and Healthcare workspaces get a beta, which stays off until a workspace admin enables it. Plus, Go and Free are not on the list.
Malaysia and Singapore are both supported ChatGPT countries, and neither is excluded for Pro. Rollout is gradual, and OpenAI says "Access may take several days to reach your account". Dots are not yet available to users under 18.
Prices, as of 8 October 2026:
- First dot: included in Pro or Business Premium at no extra cost.
- Pro: from MYR 420 a month as shown to visitors in Malaysia on ChatGPT pricing; OpenAI lists the tiers at US$100, US$200 or US$500 a month (Pro tiers).
- Business Premium seat: US$100 a month billed annually, or US$125 billed monthly. Viewed from Malaysia, the page showed MYR 390 a month. Business is for teams of 2 to 200.
Usage limits, still settling:
- For the first month after the 29 September launch, dots usage won't count toward eligible plan allowances. OpenAI will publish usage terms for each plan after that (release notes).
- Tasks your dot starts in Codex or ChatGPT Work, two other OpenAI tools, still count toward your usage limits as usual.
- More dots per person are planned, but OpenAI has not published that pricing.
For every ChatGPT plan in ringgit, see our guide to ChatGPT plans in Malaysia.
How to set up your first dot
Setting up takes a computer, not a phone. OpenAI says you cannot currently create a dot on mobile, and mobile web is not supported.
- Create it on desktop. Open the ChatGPT desktop app or ChatGPT in a desktop browser, and follow the onboarding prompts. You can name your dot along the way.
- Connect only the apps it needs. Review each app's permissions before you connect it. Then let the dot introduce itself.
- Set your Custom Rules before the first real task. The next section covers the four behaviours.
- Add messaging channels from desktop. Slack is connected from desktop, in the app or a desktop browser, and OpenAI says you can also message your dot in Microsoft Teams. Texting is a US-only beta for Pro users at launch.
- Then use it anywhere. After set-up, you can message your dot from the ChatGPT mobile app.
How you stay in control of a dot
A dot can act for you, so its controls matter more than its features. There are four worth knowing.
Custom Rules. "Dots start with built-in rules for when to act independently and when to ask for approval." Custom Rules change that per action, with four behaviours:
- take action without asking
- take action if pre-approved, meaning you explicitly asked for it in your prompt
- ask before taking action
- hand off to you
You can add boundaries such as telling your dot never to send emails. You cannot use a rule to switch off core safety requirements or auto-review.
Auto-review. Before certain actions, auto-review checks the plan against your instructions and Custom Rules, as well as OpenAI's safety requirements.
Before your dot sends an email, for example, it checks the recipient and message to catch a wrong address or information you did not mean to share. For supported sign-ins, a dot can use saved passwords without exposing them to the model.
The most sensitive actions, like changing a password or transferring money, always come back to you.
Read-only research on its own initiative. When a dot looks for ways to help without being asked, OpenAI calls it "proactive research". Those research tools are read-only: they cannot send messages, change app content or control a browser or computer. Tasks you assign or schedule follow your Custom Rules and auto-review instead.
Activity View. In the desktop app, Activity View shows your dot's ongoing and delegated tasks and the steps it has taken. You can redirect your dot at any point, or pause it.
OpenAI is plain about the limits. Prompt injection means hidden instructions planted in a web page, email or document. OpenAI's protections "help reduce the risk of malicious instructions causing an unwanted action, but they do not eliminate it". And: "Dots can still make mistakes, so always review consequential work."
Before you turn dots on for a team: a governance and PDPA checklist
Treat this as a starting point for your own policy, not legal advice. Work through it with whoever owns data protection in your company.
Data access
- Choose the workspace on purpose. OpenAI does not use ChatGPT Business, Enterprise or Edu workspace data to train its models by default. On a personal Pro plan, each user's "Improve the model for everyone" setting decides.
- Connect the least data that does the job. A dot can review connected apps proactively and form memories from them, even when nobody asked. Do not connect an HR drive or customer database just in case.
- Disconnecting is not deleting. Disconnecting an app leaves what the dot already obtained; only deleting (resetting) the dot removes it. Individual dot memories cannot be viewed or deleted yet.
- Treat a dot's memory as a store of personal data. Today, removing one customer's details means resetting the whole dot.
App permissions
- Audit existing ChatGPT connections first. Plugin permissions are shared across dots, ChatGPT, ChatGPT Work and Codex. Apps a staff member has already connected may be available to their dot, within the permissions granted.
- Agree an approved-apps list before people set dots up, including the apps that are off limits.
Approvals
- Write Custom Rules for anything external. Until you trust the pattern, set sending emails, posting messages and purchases to "ask before taking action".
- Be specific when you pre-approve. Approving one message does not give a dot ongoing permission to contact people, and advance approval stays limited to what you authorised.
Logging and incidents
- Put Activity View in someone's routine. Check it the way you would check a new hire's work.
- File consequential approvals yourself. If a dot sends a customer quotation, keep the approved version where your team keeps records.
- Tell staff what not to share. OpenAI says human review may occur in limited circumstances, including safety cases, even with model improvement turned off.
Malaysia and Singapore data protection
- Malaysia: the Personal Data Protection Act (PDPA) Security Principle asks you to "take practical steps to protect the personal data from any loss, misuse, modification". Data processors are bound by it too.
- Malaysia, if something goes wrong: section 12B, added by the 2024 amendment, requires the data controller (the business that decides how the data is used) to notify the Personal Data Protection Commissioner of a personal data breach "as soon as practicable". Agree now who decides whether a dot's mistake counts. Our guide to section 12B breach notification covers the duty in more detail.
- Singapore: the PDPC's Protection Obligation requires "reasonable security arrangements" for personal data you hold. If a breach likely results in significant harm to individuals, or is of significant scale, the PDPC says organisations must notify it and the affected individuals as soon as practicable, and notify the PDPC no later than 3 calendar days after deciding the breach is notifiable (PDPC guide).
When a dot is enough, and when you need a custom-built agent
A dot fits work that belongs to one person and runs on apps with ChatGPT plugins, as long as someone approves what matters. Morning briefs, research, follow-up drafts and recurring checks all fit. For the wider difference between answering and doing, see chatbot vs AI agent.
In our view, it fits less well in four situations:
- The work belongs to the business, not a person. OpenAI says "Your dot works on your behalf", while specialist dots take on dedicated responsibilities inside an organisation. Specialist dots are starting as focused enterprise pilots.
- Your key system has no plugin. An in-house ERP or a local accounting system may need a custom connection before any agent can use it.
- Every run must follow fixed, auditable steps. A purpose-built workflow with a log you own is easier to test and sign off.
- The agent serves your customers directly. A dot is built to work on your behalf and report back to you. For an agent that answers customers on your own channels, with your own logs and escalation, a purpose-built agent is usually the better fit.
For those, look at custom-built AI agents designed around one workflow. You may well end up with both: a dot for each person's own work, and one custom agent for the process your business depends on.
FAQ
Are ChatGPT dots available in Malaysia and Singapore?
Yes, on eligible plans, as of 8 October 2026. Both are supported ChatGPT countries. Dots are on Pro outside the European Economic Area, Switzerland and the UK, and on Business Premium in all supported regions. Rollout is gradual, so access may take several days.
Do ChatGPT Plus users get dots?
Not at launch. OpenAI lists Pro and Business Premium, plus an Enterprise, Edu and Healthcare beta that an admin must enable.
How much do OpenAI dots cost?
As of 8 October 2026, your first dot is included in Pro or Business Premium at no extra cost. Pro starts from MYR 420 a month in Malaysia (US$100), and a Business Premium seat is US$100 a month billed annually. Usage terms after the first month have not been published yet.
Can a dot send emails or pay for things without asking me?
It follows OpenAI's built-in rules plus any Custom Rules you add. Custom Rules let you require approval for an action or block it. Auto-review checks emails before they are sent, and money transfers and password changes always come back to you.
Can I delete what my dot knows about my company?
Only by resetting it. Disconnecting an app does not delete what the dot already obtained, and individual memories cannot be viewed or deleted yet. A reset deletes the dot along with its saved memories and scheduled tasks.
How Anchor Sprint can help
Anchor Sprint is an OpenAI Select Partner. We can help you roll out ChatGPT in Malaysia or Singapore and set up your first dots with the Custom Rules and review routine above. We also train the people who use them, and that ChatGPT training is HRD Corp claimable.
OpenAI Select Partner di Malaysia dan Singapura
Planning a ChatGPT or dots rollout?
Anchor Sprint ialah OpenAI Select Partner. Kami membantu syarikat di Malaysia dan Singapura melaksanakan ChatGPT, menyediakan ejen dengan selamat melalui peraturan kelulusan dan rutin semakan, serta melatih orang yang menggunakannya.
- ChatGPT rollout for your team in Malaysia or Singapore
- Agents set up with approval rules and a review routine
- Latihan pasukan, boleh dituntut HRD Corp di Malaysia
Sources
- OpenAI, Introducing dots, 29 September 2026.
- OpenAI, DevDay 2026 Recap, 29 September 2026.
- OpenAI Help Center, Getting started with your dot, checked 8 October 2026.
- OpenAI Help Center, Dots privacy, security, and safety FAQs, checked 8 October 2026.
- OpenAI Help Center, ChatGPT release notes, checked 8 October 2026.
- OpenAI Help Center, ChatGPT supported countries, checked 8 October 2026.
- OpenAI Help Center, About ChatGPT Pro tiers, checked 8 October 2026.
- OpenAI, Business pricing, viewed from Malaysia on 8 October 2026.
- Malaysia, Personal Data Protection Act 2010 (Act 709) and Personal Data Protection (Amendment) Act 2024 (Act A1727).
- PDPC Singapore, Data Protection Obligations.