OpenAI has launched Operator, its first major AI agent designed to perform tasks on its own. Announced on January 23, 2025, by CEO Sam Altman, it’s currently available to U.S. users with a $200 ChatGPT Pro subscription. It will be rolled out to Plus, Team, and Enterprise tiers soon, with international access coming later.
Operator works through a web browser, completing tasks like booking travel, making reservations, and shopping. Users see a dedicated browser window where the AI shows its actions. Users can still take control of the screen if needed.
Powered by OpenAI’s Computer-Using Agent (CUA) model, Operator mimics human actions on websites. It can click buttons, navigate menus, and fill out forms. OpenAI has worked with companies like DoorDash, eBay, and Uber to ensure it follows their rules.
Though promising, the CUA model isn’t perfect. It struggles with complex tasks like making detailed slideshows or managing calendars. For some tasks, like banking, users will need to supervise and step in when necessary.
Operator has some limitations. There are daily task limits, and it may freeze on complicated web pages, requiring user help. Certain tasks like sending emails or deleting calendar events are restricted for now.
Despite these challenges, Operator is a big leap toward AI agents that can do much more than virtual assistants like Siri or Alexa. OpenAI is taking precautions to prevent misuse, such as phishing scams, ensuring the AI is safe to use. This marks a significant step in the evolution of AI, with the potential to change how we interact with the internet.
Reference: OpenAI
Comments