OpenAI has unveiled dots – a new class of always-on AI agents that work on their own cloud computer and pursue the user's goals around the clock.
A new era of autonomous AI work
Remember the days when artificial intelligence was just a chat window where we typed commands and waited for a response? OpenAI is signaling a change to that approach. As the company states in its announcement of September 29, 2026, it is introducing dots – agents built on the GPT-6 Astra model that work toward the user's goals even when we are not chatting with them.
According to OpenAI, dots learn from feedback and from the user's preferences, so over time they get better at understanding how you work.
How do dots work under the hood?
Each dot works on its own cloud computer, with its own browser and access to the apps the user connects. A plugin system lets it reach more than 4,000 apps. In OpenAI's examples, dots also handle programming tasks, such as turning recurring customer feedback into tested pull requests.
You can talk to a dot in ChatGPT (on desktop, in the browser and in the mobile app), in Slack and in Microsoft Teams; according to reports, a limited beta for Pro users in the US also offers contact by SMS. While a task is running, you can watch the dot's activity on its cloud computer, and with the user's permission a dot can also use the user's own computer.
Proactive research and security
When you are not actively working with your dot, it can carry out "proactive research" over read-only connections – for example, noticing a missed task such as unsent invoices. In this mode it does not send messages or change content without the user's permission.
Control comes from rules: the user can allow specific actions, require approval for them, or block them, while a separate Auto-review mechanism checks planned actions against the user's instructions and safety requirements. Changing passwords, transferring money, permanently deleting data and installing unrecognized software always require confirmation. Dots are also subject to the same permissions and controls as ChatGPT.
Market availability
Dots are reaching users gradually: the first dot is included in the Pro and Business Premium plans, while Enterprise customers (including eligible Edu and Healthcare organizations) can use the beta once an administrator enables it. OpenAI says more dots will be added in the future. The company is also working with Microsoft to integrate specialist dots with the governance and security controls of Microsoft Agent 365.
Comments