OpenAI’s Operator Lets ChatGPT Use the Web for You

OpenAI Letting some users use a new ChatGPT feature that uses it Artificial intelligence Manage a web browser to book trips, buy groceries, shop for bargains and do many other online tasks.

The new tool, called Operator, is an AI agent: it relies on an AI model trained on both text and images to interpret commands and determine how to use a web browser to execute them. OpenAI claims that it has the potential to automate many daily tasks and workday tasks.

OpenAI’s operator follows rival releases by both Google and anthropological, which is more than displayed Able to use the web. AI agents Widely seen as the next evolutionary stage Chatbots are catching on for AI, and many companies have jumped on the hype train by documenting them. In most cases, these are very limited in their capabilities and use a language model to automate things normally done with simple software.

“AI is evolving from a tool that can answer your questions to one that is capable of taking steps in the world to manage complex, multistep workflows,” said Peter Wellinder, VP of Product at OpenAI. “We’ll see a lot of impact on human productivity—but also the quality of work that people are able to accomplish.”

OpenAI acknowledges that giving ChatGPT access to a web browser introduces new risks, and it says operators can sometimes misbehave. It says it has implemented various new safeguards and plans to gradually enhance operator capabilities.

The plan is to learn from how people use the tool, said Wellinder and Yash Kumar, product and engineering leads for Computer Using Agents at OpenAI. They acknowledge that the tool can lead to unwanted bookings or purchases but add that a lot of work has gone into making sure it asks before doing anything risky. “It will come back to me and ask for confirmation before taking action that may be irreversible,” says Kumar.

OpenAI today released a new “system card” that outlines the issues the operator may encounter. This includes the possibility of it misunderstanding commands or diverging from what the user asks; misuse by users; or be targeted by cybercriminals.

“It also creates an incredible amount of security challenges,” says Kumar. “Because your attack vector area and your risk vector area increases quite significantly.”

The operator will initially be available as a “research preview” to ChatGPT users with a Pro account, priced at $200 per month. The company said it plans to gradually expand access while rolling out the tool, as it will inevitably make mistakes along the way.

In several demonstrations, the operator showed the potential for AI to take a more active role as a web assistant. The tool consists of a remote web browser and a chat window to communicate with the user.

At WIRED’s request, the operator was asked to book an Amtrak train trip from New Haven, Connecticut to Washington, DC. It went to the correct website and entered the required information correctly to bring up the schedule, then asked for further instructions. If a user logs into the Amtrak website or a browser profile with stored credit card information, the operator will be able to go ahead and book a ticket — though it’s designed to ask for permission first.

Kumar asked the operator to book a table at Beretta, a restaurant in San Francisco. The program went to the OpenTable website, found the right restaurant, and looked at the availability before asking what to do next. OpenAI says it has partnered with several popular sites, including OpenTable, to ensure operators run smoothly on them.

The new tool is based on OpenAI’s GPT-4o AI model, which can understand a browser and web pages and converse in typed text. The tool includes additional training to help you understand how to perform tasks online. OpenAI will make its use of computers available to agents through its API.