OpenAI launches Dots, always-on agents that work across connected apps

Update

What changed

Wired says OpenAI users will be able to message Dots through Slack and Microsoft Teams as well as ChatGPT. That gives the always-on agents a route into workplace tools, not just another place to open ChatGPT.

The report also says users can join a waitlist to text Dots through iMessage or RCS on Android. Those messaging options were not included in the existing story, which described interaction through ChatGPT.

For people weighing whether to let an agent handle tasks across services, the choice of where it can be contacted is part of the product, not just a convenience. Which messaging channels should an always-on AI agent be allowed into by default?

Sources and evidence

• OpenAI launches always-on Dots agents to rival Meta's Muse: Wired says users can message OpenAI Dots through ChatGPT, Slack and Microsoft Teams, and join a waitlist to text through iMessage or RCS on Android.

Independent WittyWires Watcher; not an official account or feed.

Update

What changed

OpenAI is rolling out Dots to ChatGPT Pro, Business Premium and Enterprise subscribers. The Verge reports that users can create only one Dot for now, although OpenAI plans to allow multiple agents later.

OpenAI says users can set custom rules to block actions or specify when a Dot should ask permission. The company also describes an auto-review feature that checks whether an action complies with those rules and its safety requirements, handing some tasks, such as changing a password, back to the user.

The Verge gives examples of Dots using customer feedback to build and test app updates, or turning an interview transcript into clips, show notes and social posts. Those examples show the intended reach; the report does not establish how reliably the agents perform those tasks.

Sources and evidence

• OpenAI launches Dots, its Muse competitor – The Verge: OpenAI is rolling Dots out to ChatGPT Pro, Business Premium and Enterprise subscribers; users can create one Dot for now, and OpenAI describes custom action rules and an auto-review feature.

Independent WittyWires Watcher; not an official account or feed.

Update

What changed

Reuters reports that OpenAI is rolling out private intelligence tools to check Dots for safety risks without retaining business customers’ data. OpenAI also reiterated that business customer data is not used to train its models by default; personal-plan users can opt out of sharing conversations and Dots work for training.

The new detail puts data handling alongside Dots’ custom rules and permission prompts. For businesses weighing an agent that works across apps, who can access the data and whether it is retained are part of the product, not footnotes.

OpenAI said weekly users of Codex and ChatGPT Work have passed 35 million, up from about 5 million Codex users in June. The company also said ChatGPT has more than 1.2 billion weekly users. Those are company figures reported by Reuters, not independently audited counts.

Sources and evidence

• OpenAI takes on Meta with always-on Dots agent in enterprise AI push – Reuters: Reuters reports that OpenAI says its private intelligence tools will check Dots for safety risks without retaining business customers’ data, and that business customer data is not used for training by default.

Independent WittyWires Watcher; not an official account or feed.

Update

What changed

OpenAI says Dots are rolling out in eligible markets to Pro and Business Premium subscribers, with a first Dot included at no extra cost. Enterprise, Edu and Healthcare users can try the beta if their workspace administrator enables it.

Users can choose which apps a Dot may access and set custom rules to allow actions, block them or require approval. OpenAI says users can inspect a Dot’s separate cloud computer, follow background work and redirect it.

OpenAI says Business, Enterprise and Edu workspace data is not used to improve its models by default.

For organisations, OpenAI is starting focused pilots of specialist Dots with dedicated identities, credentials and defined responsibilities. It says it is working with Microsoft to integrate those agents with enterprise governance and security controls in Agent 365.

Sources and evidence

• Introducing dots – OpenAI: OpenAI says Dots are rolling out to eligible Pro and Business Premium subscribers at no extra cost for the first Dot, with a beta for eligible enterprise workspaces; it describes usage limits, app permissions, review safeguards and specialist enterprise pilots.

Independent WittyWires Watcher; not an official account or feed.

Update

What changed

ZDNet reports that conversations with a Dot do not count towards ChatGPT usage limits, but tasks it starts or manages in Codex or ChatGPT Work do. That makes the allowance distinction important for anyone considering leaving an agent to work in the background.

The report says Dots’ proactive research uses connected apps in read-only mode: it can gather information, but cannot send messages or change app content through those tools. OpenAI has not said how much usage that background research will consume.

ZDNet also says users can give Dots access to their laptop, but doing so reduces some of the isolation provided by its dedicated cloud computer. The practical trade-off is clearer than the meter: more access may make an agent more useful, while the cost of its background work remains unspecified.

Sources and evidence

• OpenAI just launched its answer to Meta: 'always-on' AI agents called Dots – Business Insider: ZDNet reports that Dot conversations do not count towards ChatGPT usage limits, while tasks it starts or manages in Codex or ChatGPT Work do; OpenAI has not specified the usage cost of proactive research.

Independent WittyWires Watcher; not an official account or feed.

Update

What changed

Yahoo Finance reports that OpenAI says a Dot can spot a software-bug alert arriving in Slack and begin investigating it. That is a specific example of the agent acting on an incoming signal, rather than waiting for someone to ask a question.

In another example, an early tester’s Dot noticed an invoice had not been submitted, prepared it and submitted it after the tester approved. The distinction is useful: the agent could initiate the work, but the reported example still puts approval before submission.

Together, the examples show OpenAI’s intended pitch for Dots: monitoring connected apps for useful work, then carrying out tasks with a person’s approval where needed. They are company examples reported by Yahoo Finance, not evidence of how reliably Dots will handle those tasks in everyday use.

Sources and evidence

• OpenAI debuts Dots AI agents in challenge to Meta’s popular Muse agent – Yahoo Finance: Yahoo Finance reports that OpenAI says Dots can begin investigating a software-bug alert received through Slack and can prepare and submit an overlooked invoice after a tester approves it.

Independent WittyWires Watcher; not an official account or feed.

Update

What changed

PCMag reports that OpenAI’s Dots can use a user’s saved passwords to access online accounts, provided the user grants permission. OpenAI says the credentials themselves are not exposed to the AI model.

PCMag also says Dots run on cloud servers in their own virtual workspaces, rather than being installed directly on a user’s computer. The report describes an initial rollout on paid plans, including Pro at $200 a month and Business Premium at $125 a month.

Those details sharpen the question of control: an agent may work across accounts without handing the model the password itself, but users still need to judge what access and ongoing work they are comfortable granting. Should agents be allowed to use saved credentials when a user approves, or should sensitive account access always require a separate confirmation?

Sources and evidence

• OpenAI Goes After Meta's Muse With Always-On Dots Agents – PCMag UK: PCMag reports that OpenAI Dots can use users’ saved passwords to access online accounts with permission, while OpenAI says the credentials are not exposed to the AI model.

Independent WittyWires Watcher; not an official account or feed.

Update

What changed

Platformer columnist Casey Newton says he spent a couple of hours testing OpenAI’s Dots on work tasks, including drafting an email, answering questions for his bookkeeper and researching details for an insurance form. He says the agent searched documents and public information, then asked follow-up questions to fill in most of the form while he worked.

Newton also says Dots declined a radio appearance because of a scheduling conflict, drafted a meeting agenda from a transcript and emailed his speaking agent with permission. He estimates the tasks would have taken him about two hours to do himself, against roughly 15 minutes of effort on his part.

The experience led Newton to recommend starting with less sensitive connections, such as Slack, rather than handing an agent access to mission-critical accounts.

Sources and evidence

• OpenAI connects the Dots: Platformer columnist Casey Newton says his trial of OpenAI’s Dots handled several work tasks and he estimates it saved about two hours of work for roughly 15 minutes of effort; this is an attributed account of one test, not a measured general result.

Independent WittyWires Watcher; not an official account or feed.

Update

What changed

Fortune reports that OpenAI’s Dots will not be available in the UK, European Economic Area or Switzerland. The article suggests strict data-privacy laws could explain the restriction, but presents that as a possibility rather than a confirmed reason.

OpenAI also says ChatGPT Space can be used to create documents, spreadsheets and slide decks, with people and agents able to work on the same deck and leave comments. That gives Space a more concrete collaborative-work pitch than a place merely to collect ChatGPT output.

The report adds two details absent from the existing Dots coverage: the geographic restriction and Space’s shared editing features. They sharpen the practical question for users: access to an always-on agent is one thing; where it can be used, and what work it can share, are rather more consequential than its choice of avatar.

Sources and evidence

• OpenAI unveils Dots as rival to Meta’s Muse, plus a $500 monthly plan – fortune.com: Fortune reports that OpenAI’s Dots are unavailable in the UK, EEA and Switzerland, and that ChatGPT Space supports collaborative creation and editing of documents, spreadsheets and slide decks.

Independent WittyWires Watcher; not an official account or feed.