OpenAI launches Dots to challenge Meta's Muse AI

During its DevDay keynote, OpenAI announced the launch of Dots, always-on AI assistants designed to perform background tasks across connected applications. Powered by the company's GPT-6 Astra model, these agents operate on their own cloud computers, allowing them to access web browsers and integrate with more than 4,000 supported apps. Users can communicate with their Dot through a text-based chat interface, voice calls via ChatGPT, or through integrations with Microsoft Teams and Slack.

For practitioners, Dots represent a shift toward highly autonomous workflows. Currently restricted to one agent per user, a Dot can handle complex multi-step tasks. For example, a developer can task a Dot with building and testing software updates based on customer feedback, while a creator can have it analyze interview transcripts to generate video clips, show notes, and social media posts. Users can monitor their agent's progress in real time by viewing its cloud computer or collaborate with team members in a new shared environment called ChatGPT Space.

To mitigate privacy and security risks, OpenAI has implemented built-in rules for autonomous behavior alongside an auto-review feature. Users can define custom boundaries to restrict certain actions or require explicit approval before the agent proceeds. The system is designed to automatically identify sensitive operations, such as password changes, and hand those tasks back to the human user.

The feature is rolling out to ChatGPT Pro, Business Premium, and Enterprise subscribers, and interactions do not count toward standard ChatGPT usage limits. OpenAI is also testing specialist Dots tailored for specific corporate roles, with future updates planned to let users scale an agent's speed and monthly workload capacity.