OpenAI, the developer of ChatGPT, has unveiled an artificial intelligence (AI) agent called "Operator" that performs various tasks in web browsers on behalf of users.
On Nov. 23 (local time), OpenAI previewed the AI agent "Operator," which can independently perform specific tasks in web browsers, through its website.
OpenAI introduced "Operator" as one of its first agents that can independently perform tasks for users. It explained, "When users issue commands, the agent executes them."
"Operator" can automatically execute tasks such as vacation planning, hotel and restaurant reservations, food delivery, and online shopping. For example, if a user commands, "Book accommodations for my trip to New York next week," the AI opens a dedicated browser to make the reservation.
"Operator" is powered by the "Computer Use Agent model" (CUA), which combines the capabilities of OpenAI's latest GPT-4o and advanced reasoning models. This allows the AI to view and interact within the web browser operating on a computer.
If an issue arises, it attempts to resolve it using its reasoning capabilities and will call the user if unresolved. Tasks such as financial transactions must be performed directly by the user. It does not function for sending emails or deleting calendar events.
"Operator" will initially be available for research to "ChatGPT Pro" subscribers in the U.S. who pay a monthly subscription fee of $200.
Although it is planned for release in countries outside the U.S., OpenAI noted that it may take time in Europe, where regulations are strict.
It was also mentioned that, currently, it is only available through a separate site, but it is expected to be integrated into ChatGPT in the long term.