The operator of OpenAI allows ChatGPT to use the Web for you

OpenAI allows some users to test a new ChatGPT feature that uses their artificial intelligence to use a web browser to book travel, buy groceries, hunt for deals, and do many other things online.

The new tool, called Operator, is an AI agent: it relies on an AI model trained on text and images to interpret commands and understand how to use a web browser to execute them. OpenAI claims that it has the potential to automate many of the day-to-day activities and things of the workday.

OpenAI’s operator follows rival versions from both Google and Anthropic, which have those demonstrated able to use the web. AI agents are widely seen as the next evolutionary stage for AI following chatbots, and many companies have jumped on the sales hype train. In most cases, these are very limited in their capabilities and simply use a language model to automate things normally done with regular software.

“AI is evolving from this tool that could answer your questions to one that is also able to act in the world, realizing complex, multistep workflows,” says Peter Welinder, VP of product at OpenAI. “We will see a lot of impact on the productivity of people, but also the quality of work that people are able to do.”

OpenAI admits that giving access to ChatGPT to a web browser introduces new risks, and says that the Operator can sometimes misbehave. It says it has implemented various new safeguards and plans to gradually expand the capabilities of the Operator.

Welinder and Yash Kumar, head of product and engineering for OpenAI’s computer user agent, say the plan is to learn from how people use the tool. They acknowledge that the tool could make reservations or unwanted purchases, but they add that a lot of work has gone into making sure they ask before doing anything risky. “He will come to me and ask for confirmation before taking steps that could be irreversible,” says Kumar.

OpenAI has also published a new “system charter” today that outlines the issues that may arise with the Operator. These include the potential for him to misunderstand commands or diverge from what a user is asking; to be abused by users; or to be targeted by cybercriminals.

“It also poses an incredible amount of security challenges,” says Kumar. “Because your attack vector area and your risk vector area increase very significantly.”

The operator will initially be available as a “search preview” for ChatGPT users with a Pro account, which costs $200 per month. The company says it plans to expand access while rolling out the tool slowly, because it will inevitably make some mistakes along the way.

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

At WIRED’s request, the operator was asked to book an Amtrak train trip from New Haven, Connecticut, to Washington, DC. He went to the right website and entered the necessary information correctly to press the schedule, then asked for further instructions. If a user was logged into the Amtrak site or into a browser profile with stored credit card information, the operator could go ahead and book a ticket, although it is designed to ask for permission first.

Kumar asked the Operator to reserve a table at Beretta, a restaurant in San Francisco. The program went to the OpenTable site, found the correct restaurant, and checked availability before asking what to do next. OpenAI says it has partnered with a number of popular sites, including OpenTable, to ensure that Operator works well on them.

The new tool is based on OpenAI’s GPT-4o AI model, which can perceive a browser and a web page and converse in typed text. The tool incorporates additional training designed to help you understand how to perform online tasks. OpenAI will also make its Computer Usage Agent available through its API.