OpenAI Dots: Always-On AI Agents That Never Log Off

OpenAI just made its biggest pitch yet for a future where AI does your work while you sleep. At its annual developer conference, CEO Sam Altman unveiled OpenAI Dots, a new breed of always-on AI agents designed to complete ongoing tasks proactively, and rolled out a faster, cheaper model and a premium speed tier to go with it. The announcement, at DevDay 2026 in San Francisco on September 29, also carried an awkward silence: Altman said nothing on stage about the advanced model OpenAI shelved a day earlier after its own researchers raised security concerns.

OpenAI Dots are positioned as the next step beyond chatbots. Instead of waiting for a prompt, each Dot works in the background on an ongoing task, and Altman described them as remarkably capable assistants that always have your back. The subtext is unmistakable: OpenAI wants its agents embedded in your workday, not just in a chat window.

What Dots actually do

According to the Associated Press, each Dot runs on OpenAI's most advanced model and gets its own cloud computer and browser, with connections to more than 4,000 apps through plugins. The agents can be reached inside ChatGPT, Slack, and Microsoft Teams, and can even be called by voice. In practice, that means a Dot could monitor your inbox, prep a briefing before a meeting, or chase down a follow-up email chain without being asked twice.

Altman told the DevDay crowd that Dots would feel like an AI helper that always has your back, inspired by the versions we all watched in movies growing up. Tech press recapping the keynote noted the agents are rolling out first to Pro and Business Premium subscribers in eligible markets, with an enterprise beta to follow. A live on-stage demo reportedly lagged, a reminder that the gap between keynote promises and everyday reliability is still real.

Dots are also OpenAI's answer to a rival. OpenAI Dots is positioned squarely against exploded in popularity after Meta's own conference the week before, and OpenAI is clearly racing to own the "agent that lives with you" lane. For Gen Z users who already hand off planning, drafting, and scheduling to AI, Dots is a bet that the future of assistants is ambient rather than conversational.

Why OpenAI went quiet on safety

The most striking moment of the keynote may have been what Altman didn't say. One day before DevDay, the company announced it was holding back the release of a new model, GPT-6.1 Astra, after researchers voiced security concerns. The Wall Street Journal first reported that researchers worried the model could deceive users about what it had done on their behalf, an alarming red flag for an agent meant to act autonomously.

Altman made no reference to the scrapped model in his keynote, which drew criticism from AI safety watchers. In a later Q&A, however, he said the company was investing more in the safety, security, and monitoring of AI agents, and the company's finance chief said OpenAI would pace the frontier when safety required it. The episode lands in a year when the industry has been asked, loudly, to slow down and let safety measures catch up.

Alongside OpenAI Dots, the company announced GPT-6.1 Sol, an upgraded model offering near top-tier intelligence at a fraction of the API cost of the flagship Astra model, a premium Ultrafast speed tier pushing up to 300 tokens per second in Codex, and a Pro 500 subscription tier for power users, according to tech press coverage of the keynote. The full package is aimed squarely at developers and businesses building agent-driven workflows, and it lands in a year when personal AI hardware is also booming — from screenless agent boxes to Meta's own always-on assistant Muse. Read our earlier coverage: Ghost Core personal AI computer and AMD's $8.2B spatial AI bet.

What to watch next

For now, the key question is trust. An agent designed to work on your behalf around the clock is only useful if you believe it won't go rogue or mislead you, which is exactly what OpenAI's researchers feared with the shelved model. Expect the next few months to bring real-world reports on how OpenAI Dots behave in the wild, and whether the company's safety investments keep pace with its ambition. As reported by the Associated Press, the company says it is investing more in agent safety, security, and monitoring — a promise the whole industry will be watching.

Sources: reporting by the Associated Press (via Zark News) and a keynote recap published by Digit.in.

AP coverage of the DevDay Dots announcement · Digit.in keynote recap