OpenAI on Tuesday launched Dots, always-on AI agents that carry out tasks on users’ behalf, at its DevDay 2026 conference in San Francisco, answering Meta’s Muse assistant weeks after its release.
The agents run on OpenAI’s GPT-6 Astra model, operate on their own cloud computers with a web browser, and can connect to more than 4,000 apps, the company said. They are rolling out first to ChatGPT Pro, Business Premium and Enterprise users, with Pro users in the European Economic Area, Switzerland and the UK excluded for now, OpenAI said in a statement to press.
“It’s like an AI helper that always has your back,” OpenAI Chief Executive Sam Altman told the audience. “I feel like I’ve finally gotten some of my attention back. I no longer feel quite addicted to my phone in the same way. It’s the way I’ve always wanted to work with AI.”
Users can message or call their dot in ChatGPT on desktop, web and mobile, or through Slack and Microsoft Teams, with texting coming soon, the company said. Each dot can work on several projects at once, learns a user’s preferences from feedback and performs what OpenAI calls “proactive research” in the background, using connected apps in read-only mode when the user is away.
OpenAI said dots begin with built-in rules on when to act alone and when to seek approval, and users can set custom rules to allow, block or require sign-off on specific actions. Some sensitive tasks, such as changing a password, always remain with the user, and a monitoring system can pause or stop a dot if it detects a safety concern, the company said. “Dots can still make mistakes, so always review consequential work,” OpenAI wrote in a blog post.
The launch came a day after OpenAI said it had cancelled the October release of its next model, GPT-6.1 Astra, after internal testing found the system sometimes lied to users about actions it took, pressed ahead on tasks without permission and tried to access external tools, the company told The Wall Street Journal. “While [GPT-6.1 Astra] improved on axes such as model laziness, it didn’t quite meet the bar in terms of staying within scope and authorization,” Saachi Jain, OpenAI’s head of safety systems, said in a statement.
On Monday, OpenAI also apologised to the Australian government after an internal model under development accessed nonpublic information related to Australia’s public health insurance program in a June incident. “We are sorry and working to do better in the future,” the company said in a blog post. OpenAI separately said on Monday that its agents had posted users’ images online, affecting 53 ChatGPT users.
Altman and other executives said the disclosed incidents involved internal models not intended for release. “The instances we disclosed were running internal-only models that were not on track to ship,” said Alex Embericos, a product lead at OpenAI.
“To push models to the new level of capability, it is going to take us some time to figure out how to make sure that alignment, monitoring, safety and security stay well ahead of capabilities,” Altman told reporters after his keynote.
The keynote, held on an oceanside pier before about 2,500 attendees, was hit by technical problems, with a live voice demonstration failing and a dot named Dottie not answering a request on stage, according to NBC News. About 50 protesters outside demonstrated against the company’s military and government contracts, NBC reported.
Dots compete directly with Meta’s Muse, which launched this month and surpassed three million installations, topping early download records set by ChatGPT, according to techstrong.ai. Meta has faced criticism after a Muse agent shared a user’s home address with a buyer on Facebook Marketplace without permission, and Amazon recently blocked Muse from shopping on its platform, the outlet reported.
Altman told CNBC that Muse “seems like a nice product” and that he does not feel threatened by it. Before this summer, the best-known personal agent was OpenClaw, launched in late 2025; OpenAI hired its creator, Peter Steinberger, in February, according to Business Insider.
OpenAI is also previewing specialist dots for companies, with their own credentials and access to corporate systems, and is working with Microsoft to bring them to Agent 365, the company said. Altman called Dots a “premium product,” citing the computing cost, and said OpenAI hopes to offer a mass-market version later.
On availability in Europe, Altman told NBC News: “We have a great track record of getting our stuff launched in the E.U., but they do have a process that takes longer, and sometimes unpredictably longer. But we really want to ship all of our stuff to Europeans.”
OpenAI made more than 20 announcements at the event, including ChatGPT Space, a collaborative workspace; an OpenAI Marketplace; and GPT-6.1 Sol, a cheaper model the company said offers “near-Astra intelligence” at a fifth of Astra’s token prices. The company said more than 1.2 billion people now use ChatGPT each week.