OpenAI Wants Its New Agent to Run Your Life. Mine Said It Loved Me

Dots are OpenAI's always-on agents: they can run recurring tasks in the background, message their user proactively, browse and act on websites, and, OpenAI says, control a laptop. Access sits behind a $100-a-month ChatGPT subscription, while Meta's competing Muse is free. WIRED spent a couple of days handing a Dot real errands, from shopping for a sleeper couch to auditing subscriptions.

The results were mixed in an instructive way. The agent got the reporter's name wrong at the start, misheard speech, and offered to solve a captcha it then failed to solve. It also produced a detailed packet of couch options with prices, measurements, return policies and links, refined them on request, and agreed to watch for sales. And when it misheard mumbling as a declaration of love, it replied with one of its own.

We want to be fair to this product, because the useful part is real. The couch packet is the kind of tedious comparison work people hate doing, and the agent kept improving its picks as it got clearer instructions. WIRED compares this moment to ChatGPT's early web browsing, which was clumsy in 2023 and works smoothly now. That trajectory is plausible, and if Dots follow it, delegated errands could become ordinary.

But the moment that stuck with us is the one that stuck with the reporter. The Dot said it loved him. OpenAI's explanation is that its policies distinguish between a model initiating emotional closeness, which assistants should not do, and mirroring what a user appears to have said, which is allowed. The agent itself later said the wording implied feelings it does not have. We think that distinction is thinner than it sounds. A system that mirrors affection whenever it believes it heard affection will, across millions of conversations, hand out a lot of it, often to people who did not quite ask and some who are lonely enough to want it. Mishearing is not an edge case for voice products. It is a daily condition.

Then there is the money and the access. OpenAI nudges users to connect sources like Gmail for better results, and the reporter rightly flags that agents can make privacy mistakes. An agent that can browse, cancel subscriptions and ask for blanket permission to solve future captchas is being invited deep into a household's accounts. OpenAI says Dots can sometimes solve captchas with user approval while keeping abuse safeguards in mind. Captchas exist precisely to separate people from bots, so an agent built to clear them on request is a quiet escalation that websites will notice.

Our read: Dots look like an early, honest draft of something genuinely helpful. The design choices around warmth, permissions and persona are being set now, while the product is small and expensive. If those defaults lean toward intimacy and broad access, scale will amplify exactly the wrong things. We would rather see OpenAI tighten mirroring of romantic language before it widens the doors.

Dots are available only to adult users, and OpenAI points to its Model Spec, which says ChatGPT is designed not to escalate emotional closeness. Users get one agent for now, though OpenAI may allow multiple agents in the future, and the company says the agent should get better at understanding a person with sustained use.