A normal AI chat ends when you close the window. OpenAI’s newest idea is different: an assistant that keeps working after you walk away.
On September 29, OpenAI introduced “dots,” its name for always-on AI agents that live inside ChatGPT. Meta and Anthropic have each launched something similar in recent weeks, so this is a trend worth understanding even if you never use one. Here’s what dots are, who can actually get one, what they might do for someone like you, and the questions to ask before you hand an AI the keys.
🤖 What Is a Dot?
A dot is an AI agent that runs on its own cloud computer, with its own web browser, and keeps making progress between conversations. You can message it in ChatGPT (and in Slack or Microsoft Teams), and it messages you back with progress, questions, or decisions that need you. It starts with what ChatGPT already knows about you, builds its own notes over time, and can connect to the apps you choose to allow. OpenAI says it works with more than 4,000.
One feature to know about: when you’re not actively working with it, a dot can do what OpenAI calls “proactive research,” looking through your connected apps for things that might need your attention. OpenAI says that background mode is read-only. It can’t send messages, change anything in your apps, or control your computer.
💲 Who Can Actually Use One Today?
Most of our readers can’t yet. Dots began rolling out on September 29 to ChatGPT Pro and Business Premium subscribers (Pro access excludes the European Economic Area, Switzerland, and the U.K.). Plus, Go, and Free accounts aren’t included, and OpenAI says more users are coming “soon” without giving a date. Pro plans run between $100 and $500 a month depending on the tier, so check OpenAI’s plan page for exactly which tiers include dots.
A few other limits: you set up a dot on a computer, either the ChatGPT desktop app or the desktop website, not on a phone. Your first dot is included in those plans at no extra charge, but OpenAI hasn’t yet published the usage limits that will apply after the first month. And dots aren’t available to anyone under 18.
💡 What This Means for You
If you use the free version of ChatGPT, there’s nothing you need to do. Dots are a brand-new product, aimed first at people who do a lot of work on a computer.
💡 What Could a Dot Do for Someone Like You?
OpenAI’s own examples are mostly about work: fixing software bugs, revising a product launch, keeping a sales proposal up to date. But a couple hint at everyday life. One early tester’s dot noticed he’d forgotten to invoice a publication, prepared the invoice, and sent it once he approved. And OpenAI’s launch page shows a dot finding a new time that works for everyone, then, with the user’s go-ahead, moving the meeting and preparing notes.
Based on what OpenAI says dots can do, here are some ideas that might fit your life. These are my ideas, not tested results:
- A morning look at your calendar and email, with a short summary of what needs your attention
- Draft replies to routine messages, which you approve before anything is sent
- A shared family trip plan or volunteer schedule that stays up to date as details change
- Background research on a purchase, a repair, or a trip, saved as notes for you to read
Notice what they have in common: the dot prepares, and you decide.
🛡️ The Part Worth Reading Slowly
An assistant that can read your email and use your accounts is exactly the kind of thing scammers would love to trick. Here’s what OpenAI says it built in:
- You choose which apps a dot can connect to, and you can review each app’s permissions first.
- “Custom Rules” let you decide, action by action, what the dot does on its own, what needs your approval, and what it must hand back to you.
- The most sensitive actions, such as changing a password or transferring money, require you to take over and do them yourself. Others, like permanently deleting data or installing software, may need your approval every time.
- You can review the steps your dot has taken, and pause it from its profile menu.
- OpenAI says dots include safeguards against malicious instructions, the kind of thing a scammer might hide inside an email.
OpenAI also says plainly that dots can still make mistakes and that you should review anything consequential. In OpenAI’s own simulated-work tests using the default confirmation settings, the underlying model still produced outcomes OpenAI classified as misaligned about 3% of the time, according to Tom’s Guide. And according to DataCamp’s summary of OpenAI’s documentation, deleting a dot’s saved memories means deleting the dot itself, and disconnecting an app doesn’t erase what the dot already learned from it.
⚠️ My Take
Even with these safeguards, I’d treat any always-on assistant as something to grow into slowly. If you ever try one, start with a single low-stakes app like your calendar, set everything to “ask before acting,” keep bank and investment accounts disconnected, and check its activity every day for the first few weeks. And remember that any message asking you to send money or share a password is a scam, even if it claims to come from an AI assistant. Our scam issue and privacy issue cover how to check, and we looked at ChatGPT’s bank-linking feature in our recent roundup.
🔭 Dots Aren’t Alone
Meta launched its own personal agent, Muse, on September 8. Google has one called Gemini Spark, for its top-tier Ultra subscribers, and Anthropic opened a waitlist preview of Claude Duet on October 21 for U.S. Pro and Max subscribers. Each has its own prices and limits, and all of them raise the same question: how much of your digital life do you hand over? We’ll keep watching and report back.
✅ Try This This Week
💬 Try This Prompt
You don’t need a dot to get ready for one. Paste this into any AI assistant: “I’d like to hand off a few routine tasks to an assistant. Help me list the things I do every week, like checking appointments, keeping track of bills, and answering routine emails, and for each one tell me what I should always approve myself and what is low-risk to hand off.” It’s a good way to find out what you’d actually be comfortable delegating.
📬 A Question You Might Be Asking
“Is it worth paying for?”
For most readers, not right now. Dots are brand new, they’re on OpenAI’s highest-priced plans, OpenAI hasn’t published the limits that apply after the first month, and independent reviews are just starting to appear. If you’re curious, wait to see whether dots reach cheaper plans, and try the free tools we cover in our other issues first.
Sources: OpenAI Help Center articles on dots (“Getting started with your dot” and “Dots privacy, security, and safety FAQs”); VentureBeat and DataCamp (Sept. 29, 2026); Tom’s Guide; Bernama/Anadolu (Sept. 30, 2026); Anthropic’s Claude Duet announcement (Oct. 21, 2026).
See you next week,
— Steve
Leave a Reply