OpenAI Dots Explained: Always-On Agents, GPT-6 Astra, and Who Can Use Them

OpenAI used its DevDay stage on September 29, 2026 to unveil Dots: always-on AI assistants the company frames as remarkably capable agents that can handle nearly anything you throw at them. Unlike a one-shot ChatGPT prompt or a single Computer Use session you start by hand, a Dot is meant to keep working on goals in the background, message you with updates, and ask when it hits a boundary you set.

For Philippine freelancers, remote workers, and SMEs already paying for ChatGPT Pro or Business Premium, the practical question is not whether the demos look cool. It is whether your plan and region qualify this week, what a Dot can safely do without you watching, and when you should still wait for clearer safety reviews and wider availability.

What are OpenAI Dots?

Dots are personal agentic assistants powered by GPT-6 Astra. TechCrunch reports OpenAI’s own description of them as remarkably capable, always-on agents built to handle everything. The Verge summarizes the company line as assistants that can “do nearly anything” across connected apps while learning preferences over time.

Each Dot runs on its own cloud computer with a browser, per The Verge and The Decoder. Through plugins, OpenAI says a Dot can connect to more than 4,000 apps and push several projects forward at once. Users name their Dot and pick a cartoon avatar—the bubbly persona TechCrunch compares to Meta’s Muse branding.

CEO Sam Altman, speaking at DevDay and to reporters, leaned into movie-style assistant imagery. Axios quotes him calling Dots “the real deal version of AI we’ve always imagined,” inspired by the cool assistants people have seen in films for years. The BBC quotes him saying you give your Dot a responsibility and it will get to work and keep working—language that matches OpenAI’s pitch of continuous background labor rather than chat-only help.

How you talk to a Dot (and what it can reach)

At launch, you interact through chat and voice inside ChatGPT on the web, desktop, and mobile. Dots also connect to Slack and Microsoft Teams and, OpenAI says, carry context across those channels. SMS support is described as coming soon in TechCrunch, The Verge, and The Decoder coverage.

The Decoder notes a product difference from ChatGPT Work’s Computer Use: a Dot works on its own cloud machine by default and can hand coding jobs to Codex; it only touches your laptop with permission. Demos show a desktop layout with chat on one side and the agent’s live work—slides, terminal, files—on the other, plus a “Take over” control when you want the keyboard back.

The Verge also mentions ChatGPT Space, a collaborative workspace where you and colleagues can work with a Dot together. That matters for PH remote teams that already live in shared docs and Slack threads more than solo chat windows.

Rules, auto-review, and what stays human

When it acts vs when it asks

OpenAI ships built-in rules for when a Dot may act on its own, plus Custom Rules so you can allow, require approval for, or ban specific actions. The Verge and The Decoder both describe an auto-review step that checks actions against those rules and safety requirements before account-touching or information-sharing steps go through. Sensitive tasks such as changing a password stay with the human.

When nobody is chatting with it, a Dot can run “proactive research” in the background. The Decoder stresses that this mode uses read-only tools only—no sending messages, changing content, or driving a browser or computer for write actions. OpenAI also says Dots can use saved passwords without the model seeing them, and that monitoring can pause or stop work if safety concerns appear.

OpenAI openly admits Dots can make mistakes and recommends double-checking any result with real consequences. That vendor caveat should sit next to every shiny demo in your mental model, especially before you let an agent draft invoices, push code, or message clients.

Availability, limits, and training defaults

OpenAI is rolling Dots out to ChatGPT Pro and Business Premium users in eligible markets; TechCrunch says users can launch from Codex or ChatGPT. The Verge’s same-day report also lists Enterprise among the rollout audiences. The Decoder’s regional breakdown is the most detailed: for now, Pro subscribers can use Dots outside the European Economic Area, Switzerland, and the UK; Business Premium customers get them in all supported regions; Enterprise, Edu, and Healthcare customers can try a beta once admins enable it. Setup starts in the desktop app or browser, then mobile.

Philippine accounts are not covered by that EEA/Switzerland/UK Pro restriction as described, but eligibility still depends on OpenAI’s supported-region lists and your specific plan. Confirm inside your ChatGPT account rather than assuming a Manila billing address auto-unlocks the feature on day one.

On training, The Decoder reports OpenAI’s defaults: Business, Enterprise, and Edu workspaces are not trained on by default; personal-plan users choose whether Dot conversations and work may be used for training. OpenAI also says it does not train directly on a Dot’s background research or the notes the agent keeps for itself—again, company policy statements, not an external audit.

If you already pay ChatGPT Pro in USD from the Philippines, the peso framing is light: you are not buying a new SKU on day one for the first Dot, but future extra Dots, speed, or monthly work volume will be paid add-ons per OpenAI’s roadmap comments. Watch the card statement in dollars; convert at your bank’s rate only when comparing against other SaaS seats.

Specialist Dots, Agent 365, and the same-week model news

For organizations, OpenAI is testing specialist Dots with fixed roles, their own identities and credentials, and optionally IT-supplied computers. TechCrunch and The Decoder say OpenAI is working with Microsoft to integrate into Agent 365 security and management controls. Internal pilot themes cited by The Decoder include purchasing, invoices, email marketing, customer support, and contract management—useful signals for PH BPOs and mid-market ops teams, still early and pilot-shaped.

Same news cycle: OpenAI also introduced GPT-6.1 Sol, positioned as a cheaper model. Some coverage (The Decoder, BBC) notes GPT-6.1 Astra was held back over safety concerns from internal testing. Launch reporting consistently says Dots run on GPT-6 Astra—not the delayed GPT-6.1 Astra variant. Attribute carefully if you mix those names in a briefing slide; they are related branding in one week of headlines, not identical shipping products.

Who in the Philippines should try vs wait

Strong try-now case: ChatGPT Pro or Business Premium users in supported regions who already live in Slack or Teams, juggle research + drafting + light coding, and can supervise an agent with clear Custom Rules. Freelancers and remotes who pay Pro for heavy weekly work may benefit from Dot chats not counting against ChatGPT usage limits, while remembering Codex and ChatGPT Work tasks the Dot starts still burn the usual quotas.

Wait or sample carefully: Anyone who needs write access to money movement, client inboxes, or production deploys without a human gate; teams in regions still blocked on Pro; orgs that want specialist Dots and Agent 365 controls before those pilots mature; users who prefer Muse-style consumer agents but are uneasy after Muse safety anecdotes The Verge references in the same cycle.

Check before you plan a rollout: Open your ChatGPT desktop or web account, confirm Pro/Business Premium status, and look for Dot setup. Europe’s Pro restriction does not automatically describe PH, but OpenAI’s “eligible markets” language still applies—treat the in-product flag as source of truth.

OpenAI’s long-term vision, stated in its announcement and echoed by TechCrunch, is teams of Dots working together on your behalf. Until that ships—and until independent testers pressure-test auto-review and proactive modes—treat the first Dot as a capable junior teammate with a cloud laptop, not as autopilot for your business.