OpenAI launches Dots, always-on AI agents powered by GPT-6

OpenAI has launched Dots, a new always-on AI agent powered by its GPT-6 Astra model. The rollout for ChatGPT Pro, Business Premium, and Enterprise customers began Tuesday at the DevDay 2026 event in San Francisco, with Pro tier access starting immediately for a $100 monthly subscription. Dots represent a shift from reactive chatbots to proactive digital coworkers that attempt to complete assigned tasks over long periods.

Unlike single-turn chatbots, Dots have an always-on design and constantly crawl the web. They pull context from connected apps, and user context is shared with them across modes. A high-powered AI agent, each Dot can tap a cloud computer, browse the web, and plug into more than 4,000 apps. They learn more about user preferences over time. OpenAI also unveiled its latest model, GPT-6.1 Sol, alongside Dots.

How users interact with Dots

Users can message Dots through ChatGPT, Slack, and Microsoft Teams, interacting via chat or voice on web, desktop, or mobile apps. Text messaging integration for iMessage and RCS on Android is on the way, with Pro users able to join a waitlist. For now, users get one Dot, which is depicted as a cute blob that can be personalized.

Dots complete tasks on a user's behalf. Users can assign multistep projects to them. In a demo, a user's Dot helped plan dinner. The Dot saw the user was working through dinner via a calendar and messaged two GrubHub food delivery options with pricing. The user followed up with a food pick and desired order time. In another demo, a user collaborated with a Dot to launch a new website. A Dot can ship app updates, clip interview highlights, and ping the user with questions and status updates. OpenAI says Dots follow built-in as well as custom rules.

Availability and rollout

The Dots rollout starts today for ChatGPT Pro subscribers. OpenAI often launches new features first for this tier before a later rollout to a larger subset of users. The feature launched for Pro, Business Premium, and Enterprise customers on Tuesday.

Enterprise focus and safety features

The Dots launch is part of OpenAI's effort to attract more enterprise clients. At DevDay, CEO Sam Altman announced specialist Dots for enterprise customers. He said they are "specialist Dots for enterprise customers tuned for specific tasks, like accounting, email marketing, and legal analysis." This adds to the trend of AI agent coworkers in professional workflows.

OpenAI built safety guardrails into Dots. They are designed to get explicit approval before sensitive actions like installing software or changing a password. Users can enable a Custom Rules tool to set boundaries, which can block unwanted actions or require direct permission. As users connect data and use agents, security and privacy risks should be considered. Automating life aspects involves trusting tools won't share unintended information. Users could be tricked by online adversaries into unwanted actions. Exercise caution when interacting with Dots and connecting pre-existing app data. If an OpenAI account allows model training on data, this setting carries over to agent interactions. OpenAI says Dots auto-review their own actions for safety and won't count against ChatGPT usage limits.

OpenAI is expected to eventually allow control of multiple agents.