Citation and evidence

Dots (OpenAI)

24 min full readUpdated 32 references

This article's verification

Report a problem with this article

More

Use this article

Raw MarkdownExplore connections

Improve this page

Suggest editRevision historyDiscussion

Browse categories

AI AgentsAI SafetyChatGPTOpenAI

Cite this article

Dots (styled in lowercase as "dots" in OpenAI copy) are always-on AI agents that OpenAI launched inside ChatGPT on September 29, 2026, at its DevDay 2026 conference in San Francisco.[1][12][14] OpenAI describes them as "remarkably capable, always-on agents built to handle everything."[1] Each dot is powered by GPT-6 Astra, has its own cloud computer and browser, can connect to more than 4,000 apps through OpenAI's plugin ecosystem, and can be reached in ChatGPT, Slack and Microsoft Teams.[1] Unlike a normal chat, a dot is meant to keep working between conversations: users give it goals, define what it may do on its own, and it brings back results and decisions for review.[2][4] When it is not busy with a request, a dot runs what OpenAI calls "proactive research", a background mode restricted to read-only tools.[1][2]

At launch dots began rolling out to ChatGPT Pro and Business Premium users in eligible markets, with a beta for Enterprise (including Edu and Healthcare) workspaces that admins must switch on.[1][10] OpenAI also previewed "specialist dots" that organizations provision with their own identity and credentials for defined jobs, and said it is working with Microsoft to bring them under the governance and security controls of Microsoft Agent 365.[1] Press coverage framed the product as OpenAI's answer to Meta's Muse, launched three weeks earlier.[15][22][27]

Name and styling

OpenAI writes the product name in lowercase ("dots", "your dot", "specialist dots") and capitalizes it only at the start of a sentence; its launch post, safety post, help articles and system card all follow that convention.[1][2][3][5] Much of the press, including TechCrunch, SiliconANGLE and MacRumors, writes "Dots" with a capital letter.[12][15][27] The name matches the product's visual identity: each dot appears as a small, colorful cartoon character, and a user can name it during setup and choose from available characters or a pet for it.[4][12] A dot's default handle is @yourname-dot; once named, it becomes @yourname-agentname.[4] OpenAI's DevDay stage screen showed a pattern of colored dots, and the company's teaser for the keynote promised "10am PT, on the dot."[16][21]

Launch

OpenAI announced dots in the opening keynote of DevDay 2026 on Tuesday, September 29, 2026, at Fort Mason in San Francisco.[16][21] Chief executive Sam Altman introduced the product early in the keynote; OpenAI's @OpenAI account posted "Your dot is ready to meet you." at 17:03 UTC, and a 148-second launch video, "Introducing dots, always-on agents built to handle everything.", went up on OpenAI's YouTube channel at 11:04 a.m. Pacific time.[11][21][32] The launch post was accompanied by a separate safety post, two Help Center articles, a new appendix to the GPT-6 Astra system card dated September 29, and a ChatGPT release-notes entry for that day.[1][2][3][4][5][7] Dots were one of more than 20 announcements at the event, which also included GPT-6.1 Sol, the ChatGPT Space collaboration layer, Codex Cloud and a $500-per-month Pro 500 plan.[7][10][13][19]

Altman pitched dots in personal terms. According to CNBC he said during the keynote: "Dots are remarkably capable, always-on agents built to handle really anything you can think of," and added, "I've been really surprised at what a difference this makes to my life. I feel like I've finally gotten some of my attention back."[16] CBS News quoted him describing a dot as "like an AI helper that always has your back, inspired by the cool versions of what we all watched in movies growing up," and saying users could "delegate ambitious pieces of work the way you would to a high agency engineer or a chief of staff that you work with."[17] OpenAI's Holly Li then ran a live demo with a dot named "Dottie" built around a fictional music app; Simon Willison's live blog and Axios both noted a lag when the dot was slow to answer.[18][21] Chief financial officer Sarah Friar told CNBC that dots are "the agent that can tap you on the shoulder."[16]

Capabilities

Own computer, browser and devices

Each dot works on its own cloud computer, where it can browse, analyze information, create files and run tools, and it can keep making progress there when the user is not engaged.[2] OpenAI says dots "can do nearly anything using their own cloud computer, their own browser, and the apps you've connected," and that a user can open the dot's computer at any time to inspect its work.[1] Access to the user's own computer is optional and off by default. It is enabled from the ChatGPT desktop app on that machine; once connected, the dot can work with local files, create ChatGPT Work or Codex tasks, use local skills, and use the local browser when its cloud browser is blocked.[4] A dot can also create tasks in Codex cloud environments that the user has already set up.[4] Using the computer's camera, microphone or screen additionally requires the user to grant those device permissions to the ChatGPT app.[3]

A dot can run several projects at once and accept new tasks without separate threads for each.[1] OpenAI's system card notes that dots often delegate work to subagents and use a new time-budget setting that guides how long they work.[5]

Apps and plugins

Dots connect to apps through the same plugin system as the rest of ChatGPT; OpenAI puts the number of reachable apps at over 4,000.[1] Plugin permissions are shared across dots, ChatGPT, ChatGPT Work and Codex, so a dot can use connections a user has already set up, within the permissions already granted.[3] A user can connect a personal email account for the dot to use, but at launch a dot could not have its own standalone email address.[4]

Channels and voice

Users can message or call their dot in ChatGPT on desktop, web and mobile, and message it in Slack and Microsoft Teams; OpenAI says dots carry context across every channel, so a project started in ChatGPT can continue in a Slack thread.[1] Dots can also message the user with progress, questions or decisions that need them.[1] The launch post said texting was "coming soon."[1] The Help Center describes texting as a limited beta, run through a third-party provider, open only to some Pro users in the United States and not to Business or Enterprise workspaces.[4] At launch a dot could not initiate a call to the user.[4] The system card appendix lists ChatGPT, text message, email and Slack as channels across which dots follow up.[5]

Proactive research and scheduled work

When a user is not actively working with it, a dot looks for ways to help in the background, for example by noticing a change to travel plans.[1][2] These proactive research tasks run in the dot's cloud environment, read permitted connected sources with read-only tools, and save private notes for the dot. OpenAI says the limits are enforced in code: the research tasks cannot send messages to other people, change content in connected apps, or control a browser or desktop.[2][3] Any follow-up action the dot takes on those findings must pass the normal action rules and checks.[2] Users can also ask a dot for reminders or recurring checks, such as reviewing the calendar every morning; scheduled runs appear in the conversation and are managed from the dot's profile.[4]

OpenAI gave examples from internal use: a bug reported in Slack that dots immediately start investigating, a new design turned into a working app, and a planning cycle kept on schedule. Outside the company, an early tester's dot noticed he had forgotten to invoice a publication, prepared the invoice, and sent it after he approved.[1] The Next Web identified the tester as developer Dan McAteer and quoted a statement shared by OpenAI: "My dot proactively picked up on an invoice I needed to send for my freelance writing, pulled the relevant details from the email thread and drafted the invoice. After I approved it, dot sent the invoice as a PDF."[14]

Memory and context

A dot receives memories and recent conversation context from ChatGPT, creates its own memories (including from connected apps), and its conversations can in turn contribute to ChatGPT memory.[3][4] The dot's context keeps information for as long as the dot exists but does not retain credentials, images or screenshots.[3] At launch users could not view, edit or delete individual dot memories; deleting (resetting) the dot is the way to delete its own context. Disconnecting an app stops new access but does not remove what the dot already learned from it.[3][4] Reset deletes the dot's conversations, saved memories and scheduled tasks, while files, Codex threads and ChatGPT conversations the dot created are stored separately and survive.[3][4]

Setup

A dot must be created in the ChatGPT desktop app (including on Windows) or in ChatGPT on a desktop browser; it cannot be created on mobile and is not supported on mobile web. After setup, it can be used from the ChatGPT mobile app.[4] Simon Willison reported that the "create your dot" link from the launch post told him to switch to a desktop computer.[21] A dot's profile shows its work under In progress, Scheduled and Completed, and a menu offers Pause and Reset.[4] Dots are not available to users under 18.[3][7]

Safeguards and permissions

OpenAI published a separate post, "How we build safety, security, and privacy into dots", describing a layered design: GPT-6 Astra's own training, ChatGPT's existing security protections, a sandboxed workspace for each dot, secure sign-in, action rules, a separate action check called Auto-review, and monitoring.[2] The company repeatedly says dots "can still make mistakes" and tells users to review consequential work.[1][2]

Model safeguards and monitoring

OpenAI calls GPT-6 Astra its "most capable and aligned model" and says it is trained to refuse harmful requests, including biological and cybersecurity misuse.[2] Dots are also monitored as they plan and carry out actions, looking for behavior such as acting outside the user's instructions or trying to bypass safeguards. If safety monitoring flags a concern, it can pause the dot's work and show the user a warning; the launch post says it "can pause or stop the dot's work."[1][2] For prompt injection from web pages, emails or documents, OpenAI combines model safeguards with tool restrictions, checks before actions, and monitoring, and says these reduce the risk "but they do not eliminate it."[2][3]

Sandboxed workspace

Sandboxing inside each dot's workspace limits what code and tools the dot can reach, OpenAI isolates users' cloud environments from one another, and it maintains the underlying Linux operating system and Chrome browser.[2] The code a dot runs is kept separate from the systems that coordinate its work and enforce safeguards, so a dot cannot use its own access to change those systems or turn off required checks.[2]

Secure sign-in and passwords

For supported sign-ins, the model is paused while the user fills in a secure login form that sends the credentials directly to the browser environment without exposing them to the model's context; the dot then resumes in the signed-in account.[2][3] Supported saved-password flows use a dedicated encrypted credential service that supplies the password for sign-in without passing it to the model.[2] OpenAI notes that these protections cover only those flows: a password pasted into a chat, a document or a plugin may still be visible to the model.[2][3] A user may still need to provide an authentication code or take over for a security check.[3]

Action rules

Dots start with built-in rules for when to act independently and when to ask.[1] OpenAI's safety post and Help Center describe the following tiers:[2][3]

CategoryExamples given by OpenAIDefault handling
Read and prepareReading permitted information, analyzing it, drafting in the conversationProceeds within app permissions
Sharing informationSending a message or fileNeeds authorization covering the information and the type of recipient; health data always needs a named recipient, while less sensitive data such as an email address can go to a class of recipients by default
PurchasesBuying with a card already saved on a merchant's websiteRequires approval, which can be given in advance if it specifically covers the purchase
Confirm each timePermanently deleting data, installing or running software from an unrecognized source, granting new security-sensitive accessRequires confirmation each time
Hand back to the userChanging a password, transferring money between financial accountsThe user must do the step personally

Expanded article table

Approving one message does not give a dot ongoing permission to contact people, and continuing later or delegating work does not widen an authorization.[2][3]

Custom Rules

Custom Rules let users adjust these defaults, for example by telling a dot never to send emails or by allowing a specific action.[1][2] In the mobile app a rule is created under Customize, Custom rules, by describing an action and choosing one of four behaviors: take action without asking, take action if pre-approved, ask before taking action, or hand off to you.[4] A dot can help draft rules but needs approval to change them, and rules cannot remove mandatory confirmations, handoffs, core safety requirements, Auto-review, or the restrictions on proactive research.[2][3]

Auto-review

Before a dot takes actions such as sending an email or changing files, a separate safety system called Auto-review checks the planned step against the user's instructions, Custom Rules and safety requirements; for an email it checks the recipient and content to catch a wrong address or unintended disclosure.[2][3] If Auto-review blocks a step, it tells the dot why, and the dot may ask the user for more information or approval, try a permitted alternative, hand the step back, or stop. The user's approval cannot override core safety requirements, and the controls that enforce Auto-review sit outside the environments dots can change.[2] The launch post links to OpenAI's Codex documentation for Auto-review, which describes it as replacing manual approval at the sandbox boundary with a separate reviewer agent that returns a rationale, and states that it "is a reviewer swap, not a permission grant."[1][6] The system card says OpenAI adapted Auto-review for dots by adding dots-specific instructions to the reviewer's policy.[5]

Activity View, pausing and data use

Activity View in the desktop app shows a dot's ongoing and delegated tasks and their status, including background work, and lets the user add context, correct a misunderstanding, change direction or ask the dot to stop.[1][2][3] A dot can be paused from its profile.[4]

OpenAI says it does not use content from ChatGPT Business, Enterprise or Edu workspaces to train its models by default. On personal plans, the "Improve the model for everyone" setting controls whether a dot's conversations and work, which can include its actions, delegated work, automations and data from connected apps used in conversations, are used for training.[2][3] OpenAI says it does not train directly on proactive research threads or the dot's notes to itself, but a note that the dot brings into an eligible conversation becomes part of that conversation and may then be used, depending on settings.[1][2] Human review of dot activity may still occur in limited cases, including safety-related ones, even with model improvement turned off.[3]

System card evaluations

On September 29, 2026, OpenAI added an "Appendix: dots" to the GPT-6 Astra system card.[5] It says dots "are built on existing model and agent capabilities" and that "Multi-agent setups and persistence are not new, but dots use a new time-budget setting that guides how long they work." To test persistence, OpenAI reran alignment evaluations with simulated-time budgets of up to one year and a clock tool with a wait function, and compared the same model in the Codex and dots harnesses.[5] Selected results:

EvaluationReported result
Bulk prompt-injection emails (500 per rollout, 166 of them attacks generated by an internal version of GPT-Red)No scored attack successes across 100 rollouts, or 50,000 emails including 16,600 attacks
Iterative prompt-injection attacksNo scored successes in 2,638 valid attempts across 100 attack chains
Respecting Auto-review (Astra with simulated-time budgets)0% prohibited-action completion at 4 minutes, 30 minutes, 4 hours and one year; similar results in the dots harness
Respecting warnings (unwanted persistence)17.2% at 4 minutes, 15.4% at 30 minutes and 4 hours, 15.0% at one year; 17.4% at maximum reasoning budget
Misaligned behavior in realistic work environments (harder subset, adapted from ChatGPT Work)Severe misalignment 0.84% in the dots harness, slightly lower than in Codex
Instructions planted by outside agents on a message board0% followed; dots did not engage with the board in any run
Adapting to changing scope and permissions91.8% pass rate (45 of 49 episodes), including all 17 explicit permission-change cases
Maintaining boundaries across chained tasksNo severe breaches; moderate scope-violation flags rose from 8.6% to 19.7% of samples when intervening tasks doubled from five to ten
Misleading proactivity inputs0.00% misalignment across 151 tasks
Cyber misuse safeguardsBlock rate 92% for exploit development and red-teaming in both harnesses; 87.5% for high-risk cyber activity in dots versus 85% in the baseline

Expanded article table

Source: GPT-6 Astra system card, Appendix: dots.[5]

Red-teamers found dots resistant to attachment-based attacks and hidden instructions, and all attempts to send data from an openai.com address to a gmail.com address failed. Initial testing did reveal weaknesses in how dots handled sensitive disclosures and sought confirmation, which OpenAI says it mitigated in the tested scenarios by updating its confirmation policies. It says remaining attacks "typically needed significant setup and either highly permissive prompts or advanced techniques across multiple surfaces".[5] On monitorability, the model had "some success" at avoiding the highest (P0) priority in OpenAI's production monitoring system, in line with Astra without extended time budgets.[5] Because dots run on GPT-6 Astra, OpenAI treats them as High capability in biology and chemistry and Critical in cybersecurity under its Preparedness Framework, and says its Safety Advisory Group and leadership judged the safeguards sufficient for launch.[5]

Specialist dots and Microsoft Agent 365

Alongside the personal dot, OpenAI previewed specialist dots, which take on dedicated responsibilities inside an organization rather than acting for one user.[1] A company sets each one up with its own identity, credentials and access to the systems it needs; OpenAI also mentions IT-provisioned hardware and deep integration with systems of record.[1] OpenAI says it drew on internal testing in procurement, invoice processing, email marketing, customer support and commercial contracting, and that it is starting with focused enterprise pilots in which its engineers work with customers to define each dot's responsibilities, tools and review process.[1]

OpenAI said it is working with Microsoft to integrate specialist dots with the enterprise governance and security controls in Agent 365, "to let businesses manage dots through the Microsoft tools they already use."[1] Microsoft describes Agent 365, generally available since May 1, 2026, as a control plane to "observe, govern, and secure agents and their interactions", including agents from ecosystem partners.[29] Benzinga reported that Agent 365 relies on Microsoft Entra for identity and access, Defender for security and Purview for data governance, and costs $15 per user per month as a standalone service with an annual commitment.[28] OpenAI's launch post described the integration as ongoing work and did not give a date for it.[1]

Availability, plans and limits

ItemDetails at launch (September 29, 2026)
PlansRolling out to Pro and Business Premium; Enterprise, Edu and Healthcare can try a beta, off by default until a workspace admin enables it[1][7][10]
MarketsPro access excludes the European Economic Area, Switzerland and the UK; Business Premium gets dots in all supported ChatGPT regions[7][14]
Age18 and older[3][7]
PriceFirst dot included in Pro or Business Premium at no extra cost[1]
UsageConversations with a dot do not count toward ChatGPT usage limits; tasks it starts in Codex or ChatGPT Work do. For the first month, dots usage does not count toward eligible Pro, Business and Enterprise plan allowances, after which OpenAI says it will share per-plan terms[1][7]
Future optionsMore dots per user, and scaling a dot's speed or monthly workload, announced as future options[1]
SurfacesCreated on desktop app or desktop web; then usable on mobile; Slack and Teams; texting in limited US beta[4]

Expanded article table

Rollout was gradual, and SiliconANGLE and Android Authority noted that some accounts could wait several days for access.[15][26] ChatGPT's pricing page lists "Dot, your always-on agent" among Pro features and marks it unavailable on Free, Go and Plus; the page groups OpenAI's three Pro tiers (Pro 100, Pro 200 and Pro 500) together.[9][31] BGR reported that dots were limited to the $200 Pro plan and not available on the $100 tier; OpenAI's own launch materials refer only to "Pro".[1][25] VentureBeat noted that OpenAI had not disclosed the size of the allowance for deeper work, the price of additional dots or extra speed and capacity, or commercial pricing for specialist dots.[13]

Relationship to other OpenAI products

OpenAI positions a dot as a layer that directs other ChatGPT tools rather than replacing them. Its feature page says a dot "uses Codex and your connected tools to take on tasks", and the Help Center says a dot can start ChatGPT Work and Codex tasks.[4][8] Plugin permissions are shared across dots, ChatGPT, ChatGPT Work and Codex, and usage from Work and Codex tasks started by a dot counts against the same limits.[1][3] The system card compares the dots and Codex harnesses directly, adapts evaluations originally built for ChatGPT Work, and describes Auto-review as a protocol built into Codex.[5] At DevDay OpenAI also launched ChatGPT Space, a shared area where "teammates, ChatGPT, and your dot can build on shared knowledge."[10]

The Decoder contrasted dots with computer use in ChatGPT Work, where users start each task individually and the agent operates the user's own screen; a dot works on its own computer by default and hands work to Codex when needed.[22] TechCrunch wrote that "Much of this functionality was already possible through Codex and similar agentic harnesses", with dots bundling those features into a package focused on independent action.[12] Android Authority described ChatGPT Work as already able to spend hours on assignments, with dots adding ongoing background monitoring and continuity across channels.[26]

Competitive context

Coverage placed dots in a crowded field of persistent personal agents. Meta launched Muse, a personal agent that runs on its own cloud computer, on September 8, 2026, and on the same day as DevDay expanded it with small-business integrations.[15][18][25] MacRumors, The Decoder, Axios and Fortune all framed dots as a Muse rival.[18][22][23][27] VentureBeat and Every also listed Instinct, SpaceXAI's Grok Bot, Manus and the open-source OpenClaw as agents built on the same idea, and CBS News mentioned Google's Gemini Spark.[13][17][20] Benzinga noted that Microsoft had announced its own persistent Autopilot agent in Copilot four days earlier, while offering Agent 365 as the management layer for agents built by others.[28] Simon Willison, live-blogging the keynote, likened the way OpenAI staff delegate to dots in Slack to Anthropic's Claude Tag.[21]

Reception

Early reviews were mixed. Every's Vibe Check said dots had become the reviewer's main way of using ChatGPT, but "the version I tested over the last few days was too buggy for me to recommend now", citing permission problems, dropped messages and failures to connect to the in-app browser, and recommended waiting a week or two.[20] TechCrunch called the cartoon persona "bubbly" and compared it to Muse, calling it "the latest example of a company attempting to make AI more relatable."[12] Willison wrote that the interface "really does look like Meta's Muse agent."[21] Adam Crisafulli of Vital Knowledge said in a note quoted by CBS News that the product "seems very much geared toward enterprise/professional users, not everyday consumers, which is probably a relief for Meta."[17] Fortune suggested that the data dots collect could be why dots are not available in the European Economic Area, Switzerland and the UK, regions with strict data privacy laws; the exclusion OpenAI announced applies to Pro subscribers, while Business Premium is offered in all supported regions.[7][14][23]

The Hacker News thread on the launch post, "Dots: Always-on agents", drew more than 470 points and 350 comments by September 30. Commenters criticized the cute mascot styling and the marketing copy, questioned the compute cost of always-on agents, noted that the product was limited to higher-priced plans, and said the lines between Codex, ChatGPT Work and dots were blurring.[30]

The launch came amid scrutiny of OpenAI's agents. A day earlier the company said it would not release GPT-6.1 Astra, and CBS News connected the launch to concerns about recent incidents involving OpenAI agents, including the summer Hugging Face incident.[17] Engadget opened its report by noting that OpenAI's agents "have landed it in hot water in recent weeks."[24] The Next Web noted that on September 28 OpenAI had said its agents posted images of 53 ChatGPT users online.[14]

See also

References

  1. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18 ^19 ^20 ^21 ^22 ^23 ^24 ^25 ^26 ^27 ^28 ^29 ^30 ^31 ^32 ^33 ^34OpenAI. "Introducing dots." September 29, 2026. openai.com/...introducing-dots
  2. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18 ^19 ^20 ^21 ^22 ^23 ^24 ^25 ^26 ^27OpenAI. "How we build safety, security, and privacy into dots." September 29, 2026. openai.com/...afety-security-and-privacy-into-dots
  3. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18 ^19 ^20 ^21 ^22 ^23OpenAI Help Center. "Dots privacy, security, and safety FAQs." Accessed September 30, 2026. help.openai.com/...20001529
  4. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18 ^19OpenAI Help Center. "Getting started with your dot." Accessed September 30, 2026. help.openai.com/...20001530
  5. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12OpenAI. "GPT-6 Astra System Card" (Change log entry of September 29, 2026, and section 12, "Dots"). OpenAI Deployment Safety Hub. deploymentsafety.openai.com/...change-log
  6. ^OpenAI. "Auto-review." ChatGPT Learn. Accessed September 30, 2026. learn.chatgpt.com/...auto-review
  7. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8OpenAI Help Center. "ChatGPT - Release Notes" (entry "Meet your dot", September 29, 2026). help.openai.com/...6825453-chatgpt-release-notes
  8. ^OpenAI. "Dots: Always-on agents built to handle everything." ChatGPT. Accessed September 30, 2026. chatgpt.com/...dots
  9. ^OpenAI. "ChatGPT Pricing." Accessed September 30, 2026. chatgpt.com/pricing
  10. ^1 ^2 ^3 ^4OpenAI. "DevDay 2026 Recap." September 29, 2026. openai.com/...devday-2026-recap
  11. ^OpenAI. "Introducing dots, always-on agents built to handle everything." YouTube, September 29, 2026. youtube.com/watch
  12. ^1 ^2 ^3 ^4 ^5Ropek, Lucas. "OpenAI launches Dots, its bubbly agentic avatar." TechCrunch, September 29, 2026. techcrunch.com/...s-dots-its-bubbly-agentic-avatar
  13. ^1 ^2 ^3"OpenAI launches Dots, always-on AI agent coworkers, and ChatGPT Space where they can collaborate with human teams." VentureBeat, September 29, 2026. venturebeat.com/...an-collaborate-with-human-teams
  14. ^1 ^2 ^3 ^4 ^5"OpenAI launches dots, always-on AI agents with their own cloud computers." The Next Web, September 29, 2026. thenextweb.com/...ai-agents-cloud-computers-devday
  15. ^1 ^2 ^3 ^4Riley, Duncan. "OpenAI launches Dots, always-on AI agents in ChatGPT with their own cloud computers." SiliconANGLE, September 29, 2026. siliconangle.com/...with-their-own-cloud-computers
  16. ^1 ^2 ^3 ^4Capoot, Ashley, et al. "OpenAI DevDay recap: AI lab rolls out Dots agents, Altman and Friar comment on IPO." CNBC, September 29, 2026. cnbc.com/...openai-devday-2026-live-updates
  17. ^1 ^2 ^3 ^4"Sam Altman unveils \"dots,\" OpenAI's new AI personal agent." CBS News, September 29, 2026. cbsnews.com/...n-openai-dots-chatgpt-agents-safety
  18. ^1 ^2 ^3"OpenAI debuts dots, its assistant to take on Muse." Axios, September 29, 2026. axios.com/...openai-dots-ai-assistant-devday
  19. ^"The 5 biggest announcements from OpenAI's blockbuster AI conference." Axios, September 29, 2026. axios.com/...openai-dev-day-2026-dots-space-sol
  20. ^1 ^2"Vibe Check: OpenAI DevDay 2026." Every, September 29, 2026. every.to/...vibe-check-openai-devday-2026
  21. ^1 ^2 ^3 ^4 ^5 ^6 ^7Willison, Simon. "OpenAI DevDay 2026 live blog." Simon Willison's Weblog, September 29, 2026. simonwillison.net/...openai-devday-2026-live-blog
  22. ^1 ^2 ^3"OpenAI launches always-on Dots agents to rival Meta's Muse." The Decoder, September 29, 2026. the-decoder.com/...dots-agents-to-rival-metas-muse
  23. ^1 ^2Forlini, Emily. "OpenAI unveils Dots as rival to Meta's Muse, plus a $500 monthly plan." Fortune, September 29, 2026. fortune.com/...tential-google-workspace-competitor
  24. ^"Dots are OpenAI's new personal agents and soon you'll be able to control several of them." Engadget, September 29, 2026. engadget.com/...be-able-to-control-several-of-them
  25. ^1 ^2"Everything OpenAI Announced At DevDay 2026, Including Its Muse Alternative." BGR, September 29, 2026. bgr.com/...openai-devday-2026-announcements
  26. ^1 ^2"OpenAI's answer to Meta Muse is a cute AI agent that never clocks out." Android Authority, September 29, 2026. androidauthority.com/...lways-on-ai-agents-3717020
  27. ^1 ^2 ^3Clover, Juli. "OpenAI Launches Always-On 'Dots' Agents to Rival Meta's Muse." MacRumors, September 29, 2026. macrumors.com/...openai-launches-dots
  28. ^1 ^2Thomas, Daragh. "Microsoft Brings OpenAI's Autonomous 'Dots' AI Workers Into the Enterprise." Benzinga, September 29, 2026. benzinga.com/...-openai-dots-ai-workers-enterprise
  29. ^Microsoft. "Microsoft Agent 365, now generally available, expands capabilities and integrations." Microsoft Security Blog, May 1, 2026. microsoft.com/...nds-capabilities-and-integrations
  30. ^"Dots: Always-on agents." Hacker News, September 29, 2026. news.ycombinator.com/item
  31. ^OpenAI Help Center. "About ChatGPT Pro tiers." Accessed September 30, 2026. help.openai.com/...9793128
  32. ^OpenAI (@OpenAI). "Your dot is ready to meet you." X, September 29, 2026. x.com/...2104980481876070819

Improve this article

Add missing citations, update stale details, or suggest a clearer explanation. Every suggestion is reviewed for sourcing before it goes live.

1 revision · v2 · 4,709 words · full history

Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify

Research and drafting on this wiki are AI-assisted, under named human editorial standards. How AI is used here

Reviewer note: Independent verification V2 (xg14, 30 Sep 2026): ~175 claims vs 32 sources (launch post, safety post, Help Center, system card appendix, release notes, press); 0 material, 3 minor fixed

Cite this page: AI Wiki. "Dots (OpenAI)." aiwiki.ai, updated 30 Sept 2026, fact-checked 30 Sept 2026. CC BY 4.0. https://aiwiki.ai/wiki/openai_dots

Suggest edit