From Aryan Vatsa | Product & Market Analysis
OpenAI Dots Reviewed: Who Gets an Always-On Agent, and Why Europe Waits
On this page
OpenAI dots are always-on agents that keep working after you close the chat, and at launch each eligible user gets exactly 1. They shipped at OpenAI DevDay 2026 on 29 September for Pro, Business Premium and Enterprise. The catch for buyers is geography. Pro subscribers in the EEA, the UK and Switzerland cannot get a dot at all, while a Business Premium seat in the same countries can.
Key takeaways
- OpenAI dots are a persistent agent with its own cloud computer, not a chat mode. Each dot runs on GPT-6 Astra, has a cloud browser, keeps notes across sessions and can follow up on its own schedule.
- The regional exclusion applies to Pro, not to business seats. OpenAI's documentation lists Pro 100, Pro 200 and Pro 500 as unavailable in the EEA, UK and Switzerland, while Business Premium is rolling out worldwide.
- Enterprise admins get 8 separate switches, and every one matters. Dots, Slack, local computer access, custom rules, cloud browser, cloud network, cloud computer use and the password manager are each controlled on their own.
- The cheapest documented route for a European team is a $125 monthly Business Premium seat. OpenAI prices it at $100 per user per month on annual billing, against $25 for a standard Business seat.
What OpenAI dots are, from the documentation
A dot is a personal agent that persists between conversations. OpenAI's own help page puts it plainly: "Powered by GPT-6 Astra, your dot lives in the cloud and has its own computer and browser." That last part is the product. Your laptop can be closed and the dot keeps working.
TechCrunch's launch report described dots as agents that "operate independent of any specific hardware or interface." The same report made a fair point that most launch coverage skipped. Much of this was already possible through Codex and similar agent harnesses. What changed is packaging, persistence and a consumer price point.
The documented capability list is broad. According to OpenAI's Meet dots page, a dot "can research, analyze data, prepare documents, and build software." It can use background agents to work on several things at once. It can also delegate parts of a request to ChatGPT Work or Codex.
It also starts with relevant ChatGPT memory and keeps its own notes. Preferences, decisions and ongoing work carry across conversations and channels. Launch coverage from 9to5Google says dots connect to over 4,000 apps, including Gmail, Google Drive and GitHub through plugins.
Dots can do what OpenAI calls proactive research. When a dot is not on a task, it reads through tools you have connected using read-only access. 9to5Google's example was a dot that noticed an invoice had not been sent and drafted one for the user's approval.
That is the feature that justifies the "always-on" label. It is also the feature a security team should read twice, because an agent that looks for work is an agent that reads data nobody asked it to read. OpenAI states that it "doesn't train models directly on proactive research or its private notes." That is a training commitment, not a retention commitment.
You create a dot from the ChatGPT desktop app or a desktop browser, then talk to it in ChatGPT, on mobile, by voice call, in Slack or in Teams. OpenAI's help page is explicit that mobile web is not supported. TechCrunch reported that text message support is coming soon. In Enterprise workspaces, the admin guide says Teams access is limited to an invite-only alpha, which is narrower than the launch coverage implied.
Which plans get OpenAI dots, and what each costs
Eligibility is the most confusing part of the launch, so here it is straight from OpenAI's documentation. Three plan families qualify. Plus does not appear in the eligibility text, and neither does a standard Business seat.
| Plan | Dots access | Region rule | Price we could verify |
|---|---|---|---|
| Pro 100, Pro 200, Pro 500 | Included, 1 dot, users over 18 | Not in the EEA, UK or Switzerland | Pro 500 at $500 a month |
| Business Premium seat | Included | Rolling out worldwide | $125 monthly or $100 annually per user |
| Enterprise, Edu, Healthcare | Beta, off by default, admin must enable | Rolling out worldwide | Contracted, not public |
| Plus and standard Business | Not listed as eligible | Not applicable | Business standard $25 monthly |
Pro 500 pricing is from Reworked's DevDay coverage. Business seat pricing is OpenAI's statement as quoted by CIO. We could not open OpenAI's pricing page directly, so the Pro 100 and Pro 200 prices are left out rather than inferred from their names.
Usage limits work differently for dots
One detail changes the cost maths. OpenAI's documentation says "Conversations with your dot don't count toward your ChatGPT usage limits." Work the dot hands off to ChatGPT Work or Codex does count, as usual.
So the real limit on a dot is not chat volume. It is how much heavy work it delegates. Reworked reports that Pro 500 carries 25 times the ChatGPT Plus allowance, and CIO reports that a Business Premium seat carries 5 times the usage of a standard seat. Neither figure is stated in dot-specific units, which is a gap buyers should raise with their account team.
OpenAI warns that "You may not see dots immediately, even if your plan is eligible." If you are piloting this for a team, budget a few days of slack before the first user can create one. We would not schedule a demo for week one.
The OpenAI dots Europe and UK exclusion, read closely
This is the question most European readers arrive with. Dots availability in the EU is not "no". It is "not on a personal plan." The documentation excludes Pro users in the European Economic Area, the United Kingdom and Switzerland, and says nothing similar about Business Premium or Enterprise.
OpenAI has not said why
We searched OpenAI's documentation and the launch coverage from TechCrunch, 9to5Google and Reworked for a stated reason. There is none. TechCrunch's report only says dots are available "in eligible markets." Anyone telling you the exact legal trigger is guessing.
Here is our reading, labelled as inference. The split falls along the line between a consumer contract and a business contract, not along a line of compute capacity or model access. If it were capacity, Business Premium would be blocked too. That points to the consumer-facing obligations, such as how a personal agent reads a private inbox and contacts other people, as the likely sticking point.
Residency settings already disable part of the product
There is one documented case where data location switches a dot capability off. The admin guide states that local computer access is unavailable for dots if enforce_residency is enabled in any cloud policy. Work Cloud stays available. If your organisation runs EU residency, expect a dot that works in its own cloud computer but not on your staff laptops. The wider residency trade-offs are covered in the comparison of EU and India AI data residency options.
The exclusion looks less unusual against the rest of the market. Meta launched its Muse personal agent on 8 September 2026, and Tech Insider reports that it launched in the US only. Personal agents that act on someone's accounts are reaching European consumers last, from more than one vendor. We think that pattern will hold into 2027.
OpenAI dots admin controls: the 8 switches that matter
For Enterprise buyers, this is the section to print. Dots in an Enterprise workspace are off by default. The main switch lives at Workspace settings, then Permissions & roles, then Permissions, then Workspace default, then Workspace capabilities. OpenAI's admin guide names each control separately.
| Setting | What it governs | What to decide first |
|---|---|---|
| Use dots (Beta) | Whether members can create a dot at all | Workspace default or a custom role for a pilot group |
| Add dots to Slack | Slack participation for members' dots | Which channels a dot should ever see |
| Allow local computer access | Dots working on a member's own machine | Needs Work Cloud enabled first; off for residency-enforced tenants |
| Use custom rules for dots | Members writing their own approval rules | If disabled, saved rules stop applying; default rules remain |
| Cloud browser use | The dot's own browser | Configured separately from network access |
| Cloud network access | Outbound network from the dot's cloud computer | Disabling it does not disable the browser |
| Cloud computer use | The dot's cloud computer, shared with ChatGPT Work | Changing it affects Work too |
| Use password manager | Saved logins the dot can reuse | The highest-risk switch on this list |
Custom roles are additive, which cuts both ways
Only workspace owners can configure custom roles. Each permission can be set to Default, On or Off. The guide states that role permissions are additive: turning a permission off in one role does not remove access granted by another. If a member sits in 2 groups, the more permissive group wins. Audit group membership before you trust a restrictive role.
One line in the admin guide deserves more attention than it got. It says that Enterprise model controls and defaults do not apply to dots. If your governance relies on restricting which models staff can use, dots sit outside that fence. Our earlier comparison of enterprise governance controls across ChatGPT, Claude and Gemini assumed model controls were the main lever. For dots, they are not.
Audit logging needs checking before you rely on it
The Compliance API is OpenAI's supported route for reviewing user messages and dot replies. The guide itself warns: "Confirm record coverage before relying on it for an audit." Compliance Logs Platform records are available for 30 days. If you need a longer trail, plan to export it. The fields worth capturing are listed in the agent audit log specification.
How dots decide when to ask you first
The approval model is where OpenAI has done its most careful work, and where buyers misread it most. Before a dot takes an action that could affect your accounts or share information, an automatic review checks it. The review weighs your instructions, permissions, custom rules and built-in safety requirements.
Users set custom rules under Settings, then Personalization, then Custom rules. Each rule takes one of four behaviours: "Take action without asking", "Take action when you say so", "Ask before taking action" or "Hand off to you". The last one exists for steps OpenAI will not let the agent do, such as changing a password.
Two documented details are reassuring. "Asking your dot to draft replies doesn't give it permission to send them." And reusing a saved login for a new sign-in always requires confirmation, whatever your rules say.
What the controls do not do
OpenAI is candid here, and we give it credit. Custom rules "are instructions your dot tries to follow, and it can make mistakes." Pausing a dot stops the current main task but not delegated tasks or future scheduled runs. Stopping work "doesn't undo completed actions." Deleting a dot does not recall messages it already sent to other people.
Read together, those sentences describe an agent whose brakes are soft. That is honest, and it is also why the admin layer matters more than the user layer. If prompt injection through a connected inbox worries you, and it should, the threat model is set out in the analysis of prompt injection against agents.
OpenAI dots against the always-on agents already on the market
Dots did not launch into an empty field. Meta's Muse arrived 3 weeks earlier, and both TechCrunch and 9to5Google framed dots as OpenAI's answer to it. The useful comparison for a buyer is not features. It is who each product is built to be bought by.
Tech Insider reports Muse pricing at $20 a month for the Power tier and $100 for Maximum, plus a free tier that still requires a payment card. It runs in a dedicated virtual machine with its own browser, a design close to dots. It launched US-only and age-gated at 18.
The difference is the admin story. We found no Muse equivalent to the 8-permission enterprise console OpenAI documents for dots. For a company buying agents for staff, that gap is decisive.
TechCrunch reported that OpenAI is working with Microsoft to integrate dots into Microsoft's Agent 365 security controls. If that lands, dots become governable inside the identity stack many enterprises already run. Until it does, treat each dot as a new non-human identity with its own credentials, using the approach in the guide to agent identity and non-human access.
| Criterion | OpenAI dots | Meta Muse |
|---|---|---|
| Launch | 29 September 2026, DevDay | 8 September 2026 |
| Where it runs | Own cloud computer and browser, optional local machine | Dedicated secure VM with its own browser |
| Regions | Pro excluded from EEA, UK, Switzerland; business plans worldwide | US only at launch |
| Admin controls | 8 documented workspace permissions plus custom roles | None documented for organisations |
| Entry price | $100 per user annually on Business Premium | Free tier with card, $20 paid tier |
| Where Muse wins | Dots cost 5 times more at the cheapest paid tier | Cheaper for an individual in the US |
Where this review is weakest
We said every post would name its own soft spots. This one has three, and the first is serious.
We have not run a dot
This is a documentation review. Nothing here measures whether a dot completes tasks well, how often its approval review fires, or how often it gets things wrong. OpenAI says plainly that a dot "can make mistakes." A buyer should weigh hands-on testing above anything in this post. A vendor-neutral checklist for that test is in the AI agent procurement checklist.
The pricing is partly second-hand
We could not open OpenAI's pricing page or its help center during research; both refused automated access. Business Premium pricing comes from OpenAI's statement as quoted by CIO. Pro 500 pricing comes from Reworked. Muse pricing comes from Tech Insider citing Meta. These are Tier 2 sources, and the citations should be upgraded once the primary pages are readable.
The Europe explanation is inference
Our consumer-versus-business reading of the exclusion fits the documented facts. It is not confirmed by OpenAI. A competing explanation, that OpenAI simply staged the rollout by plan for capacity reasons, would also fit if business plans were smaller in Europe. We think it is less likely, and we cannot rule it out.
OpenAI dots verdict by buyer type
The right answer depends almost entirely on who is paying and where they sit.
US or non-European individual on Pro. You already have a dot. Turn proactive research on only for tools where read access is low-stakes, and leave the password manager off for the first month. We would not connect a primary personal inbox until you have watched the approval review fire a few times.
European founder or small team. Pro will not get you a dot. A Business Premium seat will, at $125 a month or $100 on annual terms. That is a real cost for one person, but it buys the business data terms you would want anyway. Check the minimum seat count with OpenAI before you commit.
Enterprise IT. Enable dots for a named pilot group through a custom role, not the workspace default. Start with cloud browser on, network on, local computer off and password manager off. Then confirm Compliance API coverage for dot conversations before the pilot ends, not after.
Our overall view: dots are the most governable always-on agent on the market today, and that is a low bar. The admin console is ahead of the competition. The guarantees underneath it are softer than the console suggests.
Frequently asked questions
What are OpenAI dots?
OpenAI dots are persistent personal agents launched at OpenAI DevDay on 29 September 2026. Each dot runs on GPT-6 Astra, has its own cloud computer and browser, keeps notes across conversations and can schedule its own follow-ups. You can message it in ChatGPT, Slack or Teams. At launch each eligible user gets 1 dot, and OpenAI says more will be allowed later.
Which ChatGPT plans include dots?
OpenAI's documentation lists Pro 100, Pro 200 and Pro 500 for users over 18 outside the EEA, UK and Switzerland. Business Premium seats include dots and are rolling out worldwide. Enterprise, Edu and Healthcare workspaces get a beta that is off by default until an admin enables it. Plus and standard Business seats are not listed as eligible.
Are OpenAI dots available in the EU and UK?
Not on a personal Pro plan. OpenAI excludes Pro users in the European Economic Area, the United Kingdom and Switzerland. Business Premium and Enterprise are listed as rolling out worldwide, so a European user can get a dot through a business seat. OpenAI has not published a reason for the exclusion or a date for Pro access in those regions.
How much does an OpenAI dot cost?
There is no separate charge for a dot. It comes with an eligible plan. A ChatGPT Business Premium seat costs $125 per user per month, or $100 on annual billing, according to OpenAI's statement quoted by CIO. Pro 500 costs $500 a month. Chatting with a dot does not count toward usage limits, but work it delegates to ChatGPT Work or Codex does.
Can admins control what ChatGPT dots can do?
Yes. Enterprise admins enable dots under Workspace capabilities and can grant them through custom roles. Separate permissions cover Slack, local computer access, custom rules, cloud browser, cloud network, cloud computer use and the password manager. Role permissions are additive, so a permissive group overrides a restrictive one. Enterprise model controls do not apply to dots.
Is OpenAI dots better than Meta Muse?
For an organisation, dots are the stronger choice today because OpenAI documents an enterprise admin console and Meta does not. For a US individual, Muse is cheaper, with a $20 paid tier and a free tier. Both run in their own cloud machine with a browser. Neither is available to European consumers yet, and neither has published independent task accuracy data.
Where to start this week
If you run a European team and want to test dots, do not buy Pro. Buy one Business Premium seat on monthly billing for the person who will pilot it, and set a 30-day review date in the calendar now.
If you administer an Enterprise workspace, open Workspace capabilities today and write down the current state of all 8 permissions before anyone asks you to turn dots on. A baseline you recorded before the request is worth more than any policy drafted after it. The broader shift in how OpenAI has priced its plans is tracked in the ChatGPT plan pricing timeline.
Before you enable an agent
Map every permission an agent gets to a named owner and a review date. The agent governance maturity model shows where most teams start, and what the next step up costs.
References
- OpenAI, Meet dots, read 16 October 2026. Used for eligibility, regional exclusion, model, capabilities, usage limits and data statements.
- OpenAI, Set up dots for work, admin guide, read 16 October 2026. Used for every admin permission, residency rule, custom role behaviour and compliance note.
- OpenAI, Control your dot, read 16 October 2026. Used for custom rule options, approval review, pause, stop and delete behaviour.
- TechCrunch, OpenAI launches Dots, its bubbly agentic avatar, 29 September 2026. Used for launch framing, text messaging and the Microsoft Agent 365 work.
- 9to5Google, OpenAI launches Dots, new always-on agents you can assign tasks to, 29 September 2026. Used for the 4,000 app figure, proactive research and the 1-dot limit.
- Reworked, The biggest announcements from OpenAI DevDay, 1 October 2026. Used for Pro 500 pricing and allowance.
- CIO, OpenAI targets heavy users with premium ChatGPT Business seats, 11 August 2026. Used for Business seat pricing and usage multiple.
- Tech Insider, Meta launches Muse AI agent, September 2026. Used for Muse launch date, pricing and US-only availability.
The weakest part of this source base is pricing: OpenAI's own pricing page and help center refused automated access, so plan prices rest on press reports quoting OpenAI and Meta. Dots are in beta and the documentation is undated, so verify every eligibility statement before you buy.
Related reading