OpenAI Introduces Operator, An AI Agent That Can Perform Tasks Autonomously

OpenAI announced this Thursday a new feature for its AI chatbot called Operator, an AI agent that can perform tasks autonomously including taking control over user’s computers.

The research preview version of the tool has started rolling out for Pro users in the United States, and the startup expects to expand Operator to Plus, Team, and Enterprise users soon. The company clarified that this initial version has limitations, but it expects to evolve based on users’ feedback.

“Today we’re going to launch our first agent AI. Agents are systems that can do work for you independently when you give them a task,” said Sam Altman in a video shared by OpenAI to introduce the new product. “We think this is going to be a big trend in AI and will really impact the work people can do, how productive they can be, how creative they can be, what they can accomplish.”

During the live video presentation of Operator, OpenAI’s team demonstrated how the new tool can book a restaurant reservation or purchase groceries by conducting its own online searches, browsing suggested websites, and applying filters to tailor the results to the user’s needs. While Operator performs these tasks, it displays a window showing its progress, allowing the user to focus on other activities while it gathers the results.

According to the details on the press release, Operator is powered by Computer-Using Agent (CUA)—an interface to interact with text fields, buttons, and menus netizens usually find on websites—and GPT-4o’s vision capabilities. It doesn’t need API integrations to see or interact, it can self-correct, and it reachs out to the user to confirm or request more information—and before making critical decisions, like buying a ticket.

OpenAI clarified that they have trained the chatbot to avoid harmful tasks, like buying guns, and reduce misalignments like performing the wrong tasks.

OpenAI also recently announced its $500 billion Stargate Project in collaboration with SoftBank, Oracle, Microsoft, and the U.S. government to build AI infrastructure.