OpenAI on Thursday released a research preview of Operator, a web automation tool that uses a new AI model called Computer-Using Agent (CUA) to control computers through a visual interface. The system performs tasks by viewing and interacting with screen elements such as buttons and text fields, just like a human would.
OpenAI is expanding its ChatGPT chatbot by adding the ability to automate tasks such as vacation planning, form completion, restaurant bookings, and grocery ordering.
Operator is currently available for ChatGPT Pro plan subscribers at operator.chatgpt.com for $200 per month. The company plans to expand to Plus, Teams and Enterprise users later. OpenAI plans to integrate these features directly into ChatGPT and later release the CUA to developers via an API.
The OpenAI Operator monitors the contents of the screen while the computer is in use and simulates keyboard and mouse input to perform tasks. The Agent using the computer processes screenshots to understand the state of the computer and makes decisions about clicking, tapping, and scrolling based on the observations.
Microsoft-backed OpenAI said users can opt out of some of the company’s training data collection by disabling ChatGPT’s “Improve the model for everyone” setting, meaning Operator’s data will no longer be used to train models. The company also said that the privacy section allows users to clear all browsing data and log out of all websites “with one click.”
OpenAI Operator competes directly with an earlier version of Anthropic, the Amazon-backed artificial intelligence startup founded by former OpenAI research leaders that developed the chatbot Claude.
In October, Anthropic introduced “Computer Use,” a feature that allows its AI agents to use computers to complete complex tasks in a human-like way. Anthropic said the AI agents can interpret what’s on a computer screen, select buttons, type text, navigate websites, and perform tasks in real time across any software or web browsing.
The tool “use computers in basically the same way that we do,” Anthropic’s chief scientific officer Jared Kaplan said in an interview. He said the robot can complete a task in “tens or even hundreds of steps.”
The generative AI market, which includes OpenAI, Anthropic, Google, Amazon, Microsoft, and Meta, is expected to generate more than $1 trillion in revenue within a decade.
Google recently agreed to make a new investment of more than $1 billion in Anthropic, a source familiar with the matter confirmed. Earlier this month that Anthropik was in talks for a $2 billion funding round led by Lightspeed Venture Partners at a $60 billion valuation.
OpenAI’s latest announcement comes as the company works toward the future of artificial general intelligence. AGI is a loosely defined metric that describes an AI that is as capable as or better than human intelligence at a range of tasks. Experts are divided on whether this is achievable and how long it might take.
OpenAI says it has implemented several privacy controls. Users can opt out of having their data used to train models via ChatGPT settings, and Operator settings allow them to clear all browsing data with one click and log out of all websites at once. If a user needs to enter sensitive information such as a password or payment information, a “takeover mode” is activated and the Operator will stop collecting screenshots.
- Chocolate Day 2025: Here’s Top 5 Chocolate You Can Gift to Your Love - February 8, 2025
- ‘Ne Zha 2’ Breaks Box Office Records, Becomes Highest-Grossing Chinese Film - February 8, 2025
- Valentine’s Day Playlist: Top 10 Best Love Songs to Make Your Night Extra Special - February 8, 2025