OpenAI Dots Always-On AI Agent Explained: Price, Who Gets It and the 1-in-5 Warning

Image: Wikimedia Commons / OpenAI, PD-textlogo (trademarked logo). Illustrative image: the OpenAI logo. No freely licensed photo of a Dot itself exists — a Dot is a software agent, not a physical product — so this post uses the company mark and an illustrative robot photo, not a picture of Dots.

TL;DR — openai dots always on ai agent explained: OpenAI's Dots are persistent ChatGPT agents that keep working on your goals after you close the chat, running on their own cloud computer and connecting to more than 4,000 apps. The first Dot is included with ChatGPT Pro ($100 a month) and Business Premium, but not with Free, Go or Plus — and OpenAI's own system card warns that in chains of 10 related tasks, nearly 1 in 5 produced a moderate permission deviation, so consequential work still needs a human review.

📌 KEY HIGHLIGHTS

• OpenAI announced Dots on 29 September 2026 at its DevDay event; a TechCrunch feature on 26 October 2026 explains what the product actually is, after weeks of launch headlines.
• A Dot is one ongoing conversation, not many chats: it keeps working between conversations on its own cloud computer and browser, and reports back through ChatGPT, Slack or Microsoft Teams.
• It runs on GPT-6 Astra and connects to 4,000+ apps through OpenAI's plugin ecosystem — the same infrastructure replacing Custom GPTs.
• Price: the first Dot is included in ChatGPT Pro ($100/month) and Business Premium at no extra cost; Enterprise access is a beta an admin must switch on.
• The catch: OpenAI's GPT-6 Astra system card records a 19.7% moderate permission-deviation rate across 10-task chains — and Pro users in the EEA, Switzerland and the UK do not get Dots at launch.

What happened?

OpenAI launched Dots, its always-on assistant, at its developer day in late September 2026, and the product has spent October moving from announcement to explanation. On 26 October 2026, TechCrunch published a feature built around an interview with Alexander Embiricos, who leads OpenAI's enterprise product team, at TechCrunch Disrupt — and that interview is the clearest account yet of what Dots is for: not a new model, but a new way of packaging what OpenAI's coding tools could already do.

Embiricos's framing, as reported by TechCrunch, is that the work was in making the agent feel intuitive. Present it as a single entity with a single conversation, rather than many separate chats, and people understand it immediately. Even the name is a piece of that framing: it plays on the negative space inside OpenAI's logo, chosen, he said, to feel like something that could eventually be ubiquitous.

The timing matters for a second reason. The same week, OpenAI stopped accepting new Custom GPTs — the community of specialised chatbots built over the last two years — and pointed builders toward the plugin ecosystem instead. Dots are the agent-era replacement: where a Custom GPT answered and the session ended, a Dot persists, accumulates context and acts over time.

openai dots always on ai agent explained: what a Dot actually does

A Dot is a persistent AI agent with its own cloud computer and browser — a dedicated environment hosted by OpenAI — so it keeps working whether or not a chat window is open. Your device is not the execution environment, and local computer access is off by default; it must be explicitly enabled through the ChatGPT desktop app.

You give it work in plain language — for example, reviewing your calendar each morning and telling you what is coming up — and track it in the Dot's profile under In progress, Scheduled and Completed. A desktop Activity View shows the ongoing and delegated tasks and the steps taken, which is the audit trail for work done while you were away.

Its reach comes from connections: more than 4,000 apps through OpenAI's plugin ecosystem, plus calendar, email and computer access where you grant it. You reach the Dot itself through ChatGPT on desktop, web and mobile, or through Slack and Microsoft Teams. One practical limit: you can only create a Dot in the ChatGPT desktop app (macOS or Windows) or on desktop web — not on mobile, though you can message it from the mobile app afterwards.

Question	Answer
Model	GPT-6 Astra (OpenAI's flagship; the planned GPT-6.1 Astra was shelved on 28 September 2026 after alignment regressions, as reported by TechTimes)
Execution	Its own cloud computer and browser, hosted by OpenAI; local access off by default
App connections	4,000+ apps via the plugin ecosystem
Channels	ChatGPT (desktop, web, mobile), Slack, Microsoft Teams; outbound calls not supported at launch
First Dot price	Included in Pro ($100/month) and Business Premium at no extra cost; pricing for extra Dots not published
Who cannot get one yet	Free, Go and Plus users; Pro users in the EEA, Switzerland and the UK; users under 18

Illustrative image: Ameca, a humanoid robot (Wikimedia Commons). This is NOT a Dot — Dots are software agents with no physical form — it illustrates the always-on assistant idea this post explains.

The controls — and the 1-in-5 warning

OpenAI's answer to the obvious question — when does an always-on agent proceed, and when does it stop and ask? — is a four-layer control system, as described in reporting on the launch materials. Custom Rules let you set, per action type, whether a Dot may proceed, must ask, or is blocked; Auto-review checks consequential actions (sending a message, modifying a file, submitting a form) against your instructions and rules before they happen, and it runs outside the Dot's own execution environment so the agent cannot override it; Activity View logs everything; and credential isolation means the Dot can use logins without seeing the passwords.

There is also a hard architectural line: proactive research — the background thinking about what could help you, which TechCrunch's Embiricos interview describes as running on a cadence — is restricted to read-only access. It cannot send messages, change content in an app or take over a browser without you asking first. Gathering information and taking actions are separate by design, not just by policy.

But OpenAI's own disclosure is the number to remember. The GPT-6 Astra system card, as reported by TechTimes, found that when Dots handle 10 related tasks in sequence, the rate of moderate permission deviations reached 19.7% — carrying context from one task into an unrelated one, or editing documents the user never targeted. With 5 intervening tasks the rate was 8.6%. In boundary tests where permissions changed mid-task, 45 of 49 cases respected the intended boundaries (91.8%), with the failures in ambiguous-scope cases — which, as TechTimes notes, is exactly what real-world tasks usually are. OpenAI's own line, quoted in the launch materials, is that "dots can still make mistakes and consequential work should be reviewed."

TechCrunch's hands-on testing adds a softer warning: the proactive alerts are real but not yet timely — Embiricos himself recalled his Dot warning him about a calendar conflict three minutes before the event — and the privacy controls are not yet granular, with the app refusing, for example, to accept read-only access to cloud drives. We have not tested Dots ourselves; these are OpenAI's disclosures and reporters' tests, not independent verification by TechByCountry.

Why it matters for India readers

For Indian readers, the first thing to check is the exclusion list — and India is not on it. The Pro exclusion announced by OpenAI covers the European Economic Area, Switzerland and the UK; Business Premium is available across all supported ChatGPT regions. So an Indian Pro subscriber is in an eligible market, subject to the gradual rollout, which OpenAI says may take several days to reach an account even after eligibility.

The real barrier in India is the tier, not the map. Dots do not come with ChatGPT Free, Go or Plus — the tiers most Indian users are on. The entry point is Pro at $100 a month, and OpenAI has announced no India-specific Dots price in the sources for this post, so that dollar figure is the one to plan around. If you are on Plus and were hoping Dots would arrive as a free upgrade, it will not: this is a premium-tier product, and the first month's grace period — Dots usage not counting toward plan allowances — is an evaluation window, not a free tier.

The Custom GPTs shutdown matters here too. Indian developers and small businesses who built Custom GPTs face the same timeline as everyone else: no new Custom GPTs from 26 October 2026, pages inaccessible from 11 December 2026 (Enterprise accounts deferred to 11 February 2027). The migration path OpenAI points to is plugins and Dots — a different interaction model to rebuild for, not a setting to flip. Readers following OpenAI's cheaper-model race can also see how this fits with OpenAI's GPT-6.1 Terra, which cut API prices for computer-use work — Dots are the agent that would spend those cheaper tokens.

Should you care? ✅ Worth watching — not worth rushing

Verdict: ✅ Follow it closely if you run a business on ChatGPT — Dots are the clearest sign yet that OpenAI's future is delegation, not dialogue, and the plugin migration affects every Custom GPT builder. ⏳ But do not hand a Dot consequential work unsupervised yet: a 19.7% deviation rate on long task chains, late alerts and non-granular controls mean the human review step is load-bearing, not optional. And if you are on Free, Go or Plus, there is nothing to buy into today — wait for a tier change, not a launch.

Price, availability & alternatives

ChatGPT Pro starts at $100 a month, with higher Pro tiers at $200 and $500 reported in coverage of the launch; the first Dot is included at no extra cost on Pro and Business Premium, and conversations with a Dot do not count toward ChatGPT usage limits — but tasks it starts in Codex or ChatGPT Work do count as usual, so an agent running nightly code review or triage can consume a plan allowance with no manual prompts at all. Enterprise, Edu and Healthcare workspaces get Dots as a beta that is off by default until a workspace admin enables it.

The alternatives frame the choice: Meta's Muse and the startup Instinct are the always-on rivals TechCrunch names, and Anthropic's Claude Duet is the proactive-agent counterpart on the Claude side — TechByCountry covered Anthropic's big week, including its S-1 filing, separately. For readers who want an agent without the $100 tier, the honest answer today is that there is no free equivalent of a persistent Dot inside ChatGPT — the free route is still ordinary prompted chats, plus the phone-side AI assistants like Siri AI in iOS 27.2, which work very differently.

FAQs

What is OpenAI Dots? Dots are persistent, always-on AI agents inside ChatGPT, announced on 29 September 2026. Each Dot runs on GPT-6 Astra, has its own cloud computer and browser, connects to 4,000+ apps through plugins, and keeps working toward your goals between conversations instead of ending when the chat ends.

How much does OpenAI Dots cost? Your first Dot is included at no extra cost with ChatGPT Pro ($100 a month) or Business Premium. OpenAI has not published pricing for additional Dots. Conversations with a Dot do not count toward ChatGPT usage limits, but tasks it starts in Codex or ChatGPT Work do.

Is OpenAI Dots available in India? OpenAI's announced Pro exclusion covers only the EEA, Switzerland and the UK, and Business Premium is available across supported regions — so India is not an excluded market. Access is still gradual and may take days to reach an eligible account, and the entry tier is Pro, not Free, Go or Plus.

Can a Dot make mistakes with my data? Yes — OpenAI says so itself. Its system card records a 19.7% moderate permission-deviation rate across 10-task chains, and its launch materials state that consequential work should be reviewed. Disconnecting an app stops new access, but data already incorporated into a Dot's context is only cleared by resetting the Dot, which also deletes its history and scheduled tasks.

What happens to Custom GPTs? No new Custom GPTs can be created from 26 October 2026, and Custom GPT pages become inaccessible on 11 December 2026 (Enterprise accounts are deferred to 11 February 2027). OpenAI points builders toward the plugin ecosystem that powers Dots — a rebuild for the agent model, not a straight migration.

Takeaway: Dots turn ChatGPT from something you ask into something you delegate to — powerful, premium-priced, and still unreliable enough on long task chains that the human stays in the loop. Watch the controls mature before you hand over the keys.

Follow World Tech on TechByCountry for tomorrow's update.

📸 Get tech in the last 24 hours on Instagram — follow @_techinlast24hrs for daily posts (9 AM–9 PM) and reels at 12:30 PM & 8:30 PM.

Sources

• TechCrunch — How OpenAI turned its coding tools into dots, an always-on personal assistant (26 Oct 2026)
• TechTimes — OpenAI Dots: Always-On ChatGPT Agents Keep Working but Fail 1 in 5 Long Tasks (26 Oct 2026)
• Dev.to — OpenAI Dots, explained from the docs: the permission model of an always-on agent (OpenAI Help Center summary)