§ Claim under review · Release
"OpenAI is rolling out 'dots', a new class of always-on AI agents inside ChatGPT powered by GPT-6 Astra, each with its own cloud computer and browser, able to connect to more than 4,000 apps, remember ongoing context, and run scheduled tasks, starting with Pro users"
Verdict
Mostly accurate
Confidence
HighSummary
This one largely holds up. OpenAI really did launch a product called dots on September 29 2026 at its DevDay event, and OpenAI's own announcement page confirms the specifics: always on agents inside ChatGPT, powered by GPT-6 Astra, each with its own cloud computer, connecting to over 4,000 apps through plugins, with memory and scheduled tasks. Two details in the post are off. OpenAI says Business Premium got dots on launch day alongside Pro, not later as planned support, and an Enterprise beta was also available if an admin switched it on. The post also leaves out that Pro access excludes the European Economic Area, Switzerland, and the UK, so Pro subscribers there could not get it at launch. Worth knowing too that OpenAI limits what a dot does on its own initiative: background research is read only, and actions that affect the outside world go through user approval rules. Everything about how well dots actually work comes from OpenAI itself, and no independent testing exists yet. Status as of October 5 2026.
The readings
key figures from the evidenceframing of dots working toward user goals continuously
Why this verdict
Evidence
OpenAI's own launch page states that dots are powered by GPT-6 Astra, that they have their own cloud computer, that they learn from feedback over time, that they can work towards user goals 24/7, and that through OpenAI's plugin ecosystem they can connect to over 4,000 apps. The same page states that dots were rolling out that day in ChatGPT to Pro and Business Premium users in eligible markets, and that Enterprise users including Edu and Healthcare can try the beta when a workspace admin enables it.
OpenAI's Help Center article adds the geographic detail: Pro availability excludes the European Economic Area, Switzerland, and the UK, while Business Premium availability covers all supported ChatGPT regions. The Enterprise beta is off by default. The first dot is included in a Pro or Business Premium plan at no extra cost.
The Help Center also confirms memory and scheduling: a dot receives memories from ChatGPT and can create its own, including from connected apps; it can review connected information proactively and form memories even without a new question; and it can run reminders or recurring tasks the user has scheduled, reviewable in the dot's profile. It confirms a four-tier approval model: take action without asking, take action if pre-approved, ask before taking action, and hand off to the user.
OpenAI's own launch thread states that a dot "has its own computer and works across 4,000+ apps through ChatGPT, using the plugins you connect," and that dots will be available on web, mobile, and desktop across Pro, Business Premium, and Enterprise in eligible markets.
Findings
✓ What's accurate 9
- dots exist and launched. OpenAI published a dedicated announcement page on September 29 2026 and announced the product at DevDay 2026. This is not a rumor or a fabricated launch
- "Powered by GPT-6 Astra" is OpenAI's own wording, appearing on the launch page and in the official launch post
- "Always-on" matches OpenAI's own framing. The launch page says dots can work towards user goals 24/7, and the Help Center says a dot can keep making progress between conversations
- "Its own cloud computer" is OpenAI's own wording on the launch page
- "More than 4,000 apps" matches OpenAI's own figure. The launch page says over 4,000 apps, and the official post says 4,000+ apps through plugins. Note the mechanism: this is reachable through OpenAI's plugin ecosystem, using the plugins a user connects, not a built-in catalogue the dot has by default
- "Remember ongoing context" is supported. OpenAI's Help Center states a dot receives memories from ChatGPT and can create its own, including from connected apps
- "Run scheduled tasks" is supported. OpenAI's Help Center describes asking a dot to set a reminder or run a recurring check, with scheduled work reviewable in the dot's profile
- The post's further statement that users can set approval rules is supported. OpenAI documents a four-tier control model running from acting without asking to handing the action back to the user
- Pro users are in the first wave. That part of "starting with Pro users" is correct
≈ What's misleading 4
- Unreleased as released, in reverse: the post says Business Premium and Enterprise support is "planned." OpenAI's own launch page says dots were rolling out that same day to Pro and Business Premium users in eligible markets, and that Enterprise users including Edu and Healthcare can try the beta when a workspace admin enables it. Business Premium was not a future plan, it was live at launch and in all supported ChatGPT regions. This is an understatement rather than hype, but it still misdescribes who could use the product on day one
- The claim's "starting with Pro users" drops the geographic restriction. OpenAI's Help Center states the Pro rollout excludes the European Economic Area, Switzerland, and the UK. For a reader in those territories, "starting with Pro users" reads as an invitation to go and find the feature, and a Pro subscriber there could not get it at launch. The same sentence also omits that the Enterprise beta is off by default until an admin turns it on
- "always-on" and the list of capabilities are presented without the permission boundary OpenAI itself documents. OpenAI restricts background proactive research to read-only, meaning that while a dot is looking for useful work on its own it cannot send messages, change content in connected apps, or drive a browser or computer. Actions with outside effects run through approvals. The post mentions approval rules in its body but the headline claim frames unrestricted continuous action
- Every capability element in the claim traces to OpenAI's own announcement, with no independent testing behind any of it. The post's own "Source: OpenAI" line is accurate, and for existence and availability the vendor is the right authority, but "remarkably capable" style feature descriptions are vendor self-report and have not been independently evaluated five days after launch
? What's uncertain 4
- Whether OpenAI itself describes a dedicated browser per dot in those words. OpenAI's headline phrasing is "own cloud computer," and the browser is attested throughout coverage and implied by OpenAI's own safety restrictions, but the exact vendor sentence was not located
- How well any of the described capabilities work in practice. No independent evaluation of dots exists yet, which is expected for a product five days old
- What "eligible markets" covers beyond the three named exclusions. OpenAI's wording is open-ended and the full market list was not located
- Whether the rollout has expanded since launch. OpenAI said it plans to expand to more users soon, and the picture here is as of 2026-10-05
Sources
6 of 8 linked to recordsOpenAI, "Introducing dots," openai.com/index/introducing-dots/, dated September 29 2026
OpenAI Help Center, "Getting started with your dot," article 20001530
OpenAI official X account, launch thread of September 29 2026 (posts 2104984504133918973, 2104984505677430978, 2104984507107717331, 2104984508454121795)
OpenAI, "How we build safety, security, and privacy into dots," openai.com
TechCrunch, "OpenAI launches Dots, its bubbly agentic avatar," September 29 2026
SiliconANGLE, September 29 2026
heise online, "OpenAI Launches Dots,"
VentureBeat, MacRumors, The New Stack, BetaNews, NBC News, Tom's Guide, MediaNama