OpenAI launches Operator, an AI agent that performs tasks autonomously

OpenAI introduces Operator, an AI agent for task automation.

: OpenAI unveiled Operator, an AI agent available as a research preview in the U.S. on ChatGPT’s $200 Pro plan. Operator automates tasks like booking and shopping using the CUA model, which mimics human interactions on web pages. The agent has limitations on complex tasks and requires user supervision for sensitive tasks like banking. OpenAI collaborates with companies to ensure compliance and implements safeguards against malicious activities.

OpenAI has announced the launch of Operator, a general-purpose AI agent designed to perform tasks autonomously by taking control of a web browser. Initially available to U.S. ChatGPT Pro users, Operator is expected to reach more users across different subscription tiers and will be integrated into all ChatGPT clients.

Operator automates various tasks, such as travel bookings and online shopping, using the CUA model. This model blends the visual abilities of GPT-4o with advanced reasoning to interact with websites much like a human would. Although it promises significant automation, it requires user confirmation before completing tasks and doesn't share or store any user data.

Despite its potential, Operator has limitations in handling complex tasks and requires supervision for sensitive activities like entering credit card information. OpenAI collaborates with partners like DoorDash and Uber to align with service terms, and safety measures are in place to mitigate risks from malicious use, setting a new precedent for AI-enabled automation.