OpenAI has launched Dots, a new class of always-on AI agents designed to work across the web and connected applications rather than respond only to individual prompts. The agents are powered by the startup's GPT-6 Astra model and are being positioned as tools for completing multistep tasks.
The first rollout is available to ChatGPT Pro subscribers, whose plan costs $100 a month. OpenAI has often introduced new features to this higher subscription tier before expanding access to a larger group of users. Dots currently allow people to control one agent, while support for managing several agents at once is expected later.
How Dots work
Dots can continuously crawl the web, draw on information from linked apps and build a better understanding of a user's preferences over time. People can contact them through ChatGPT, Slack and Microsoft Teams, with user context carried between those modes.
Pro subscribers can also join a waitlist to communicate with their Dots through iMessage or RCS messaging on Android. The cross-platform design is intended to let the agents remain part of a user's existing workflow rather than confining interactions to a single chatbot window.
OpenAI demonstrated the concept with a dinner-planning example. A Dot noticed from a user's calendar that the person would be working through dinner, then sent 2 GrubHub delivery choices with the price of each option. After the user selected a meal and specified when it should be ordered, the agent could continue with the request. Another demonstration involved a user working with a Dot to launch a new website.
Enterprise plans and safeguards
OpenAI is also targeting businesses with specialist Dots announced by CEO Sam Altman at DevDay. These versions are intended for defined professional tasks, including accounting, email marketing and legal analysis. The move places Dots within a wider push to integrate AI agents into workplace processes.
The agents are designed to request explicit approval before carrying out more sensitive actions, such as installing software or changing a password. A Custom Rules tool lets users define boundaries and identify tasks that require direct permission.
Those controls do not remove the risks that come with giving software access to personal information and other applications. Users may be concerned about unintended data sharing or an online adversary manipulating an agent into taking an unwanted action. OpenAI has built safety guardrails into Dots, but the company's rollout comes as personal agents are still developing.
Dots and the personal-agent market
The launch follows growing interest in personalised AI assistants. Early adopters in Silicon Valley had already begun experimenting with tools such as OpenClaw, while Meta's Muse recently reached a broader audience and climbed smartphone app download charts.
Dots represent OpenAI's entry into that increasingly competitive space, with a focus on persistent access to user context and action across multiple services. The company's enterprise plans suggest that the same model could extend from personal organisation to specialised workplace work.
Privacy considerations
Connecting existing apps gives Dots more information to work with, but it also increases the importance of account and data settings. If an OpenAI account permits model training on user data, that setting also applies to interactions with the agents.
The launch is initially limited to Pro subscribers, while the enterprise versions are aimed at customers seeking task-specific assistance. Broader availability and the ability to control multiple Dots are expected developments, but no further rollout details are specified.
Conclusion
OpenAI's Dots combine persistent web access, connected-app context and approval controls in an effort to make AI agents useful for personal and professional tasks. Their early availability is limited to ChatGPT Pro users, with privacy safeguards remaining an important consideration as people connect more data and services.
Frequently Asked Questions
Q. What are OpenAI Dots?
OpenAI Dots are always-on AI agents designed to browse the web, use connected-app context and complete multistep tasks.
Q. Which model powers Dots?
Dots are powered by OpenAI's GPT-6 Astra model.
Q. How much does ChatGPT Pro cost for Dots access?
The initial rollout is for ChatGPT Pro subscribers, a plan priced at $100 a month.
Q. Where can users message Dots?
Users can message Dots through ChatGPT, Slack and Microsoft Teams. Pro users can join a waitlist for iMessage or RCS messaging on Android.
Q. Can Dots take sensitive actions without permission?
Dots are designed to seek explicit approval before actions such as installing software or changing a password.
Q. What are specialist Dots for enterprises?
They are versions tuned for professional tasks such as accounting, email marketing and legal analysis.
Q. What privacy issue applies to OpenAI account settings?
When an account allows model training on its data, that setting also applies to interactions with Dots.
