OpenAI CEO Sam Altman announced that 2025 will be pivotal for AI agents, tools designed to automate tasks. The organization has now introduced Operator, a general-purpose AI agent capable of controlling a web browser and performing various actions independently. Currently available to U. S. users on the $200 ChatGPT Pro subscription, Operator will eventually be accessible to those on Plus, Team, and Enterprise plans, though European users will have to wait longer for the rollout. Operator aims to automate tasks like booking travel, making restaurant reservations, and online shopping. Within its interface, users can select tasks such as shopping and travel, which activate a dedicated web browser window where the agent operates, providing explanations of its actions.

While users can maintain control of their screens, Operator utilizes its own browser for task completion. Powered by a Computer-Using Agent (CUA) model that integrates the visual skills of GPT-4o with advanced reasoning, Operator interacts with websites similarly to a human, navigating menus and filling forms. OpenAI is collaborating with businesses like DoorDash and eBay to ensure compliance with their terms of service, and the CUA model is designed to request user confirmation for significant actions, adding a layer of oversight. However, OpenAI acknowledges the current limitations of the CUA, stating it may struggle with complex tasks and requiring user supervision for sensitive activities, like banking. Safety precautions prevent Operator from independently handling tasks like email sending or deleting calendar events for security reasons, and it may stall on complicated interfaces. Despite being slower to develop AI agents than competitors, OpenAI has emphasized safety, creating systems to minimize risks of misuse. The release of Operator marks a significant step in AI capabilities, showcasing features beyond those of traditional virtual assistants and hinting at the transformative potential of AI agents in users' online experiences.