OpenAI Unveils 'Dot,' an Agent That Works on Users' Behalf

[Financial News] OpenAI has unveiled 'Dot,' an artificial intelligence (AI) agent that learns users' goals and preferences and continuously performs tasks. It acts as an AI assistant by monitoring the progress of assigned work, carrying out necessary tasks without requiring step-by-step instructions, and bringing back items that need the user's review.
At OpenAI DevDay 2026, held in San Francisco on the 29th, local time, OpenAI announced more than 20 major updates spanning Dot, ChatGPT, OpenAI Codex, AI models and developer tools.

Built on GPT-6 Astra, Dot is provided with its own cloud computer and browser. When a user assigns it a project, Dot continues working on it while the user can check its progress or change directions along the way.
Dot can connect to more than 4,000 apps through a plugin ecosystem. With the user's permission, it can also work on other devices, such as a laptop. For example, if a user assigns Dot a product launch, the agent can proactively prepare draft revisions to the launch plan and related materials when the product's scope changes, then suggest them to the user. The user, however, makes the final decision on review and approval.
Users can communicate with Dot through messages or voice on the desktop, web and mobile versions of ChatGPT. They can also call Dot to assign tasks or receive briefings on the results.
OpenAI has also introduced controls that allow users to directly manage the scope of Dot's actions. Users can set certain actions to be approved automatically or require their approval, and can block them when necessary. Actions affecting accounts or the sharing of information are subject to automatic review, while sensitive tasks such as changing a password must be performed directly by the user.
Dot will be rolled out sequentially to ChatGPT Pro and Business Premium users. Enterprise users can access the beta version once their workspace administrator activates the feature. The Pro and Business Premium plans include the first Dot at no additional cost.

OpenAI also unveiled its next-generation model, GPT-6.1 Sol, on the same day. Compared with GPT-6 Sol, the model has enhanced capabilities in agent-based coding, computer use and professional tasks. On DeepSWE v1.1, an evaluation of long-duration software development using real-world codebases, it recorded the same score as GPT-6 Astra.
OpenAI also introduced the faster 'Ultrafast' models. GPT-6 Astra Ultrafast will be available through the application programming interface (API) and select ChatGPT Work and OpenAI Codex plans, while GPT-6.1 Sol Ultrafast is scheduled for release at a later date.
OpenAI is also expanding collaboration tools that enable people and AI agents to work together. Its newly unveiled 'Space' allows multiple users, ChatGPT and Dot to work together in one place. In 'Page,' people and agents can jointly write, conduct research and create charts and images. OpenAI also plans to launch 'Slide,' which will support collaborative work.

[email protected] Choi Hye-rim Reporter