OpenAI is letting some users try out a new ChatGPT feature that uses its artificial intelligence to control a web browser to book trips, shop for groceries, hunt for bargains, and do many other online chores.
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 figure out how to use a web browser to execute them. OpenAI claims it has the potential to automate many daily tasks and errands.
OpenAI's Operator follows competing releases from both Google and Anthropic, which have demonstrated their ability to leverage the Internet. AI agents are widely seen as the next evolutionary stage for AI following chatbots, and many companies have jumped on the hype train by touting them. In most cases these are very limited in their capabilities and simply use a language model to automate things that would normally be done with regular software.
“AI is evolving from a tool that can answer your questions to a tool that can also take action in the world and execute complex, multi-step workflows,” said Peter Welinder, VP Product at OpenAI. “We're going to see a lot of impact on people's productivity, but also on the quality of work that people can accomplish.”
OpenAI admits that giving ChatGPT access to a web browser introduces new risks, and says Operator can sometimes misbehave. It says it has implemented several new security measures and plans to gradually expand Operator's capabilities.
Welinder and Yash Kumar, product and engineering lead for OpenAI's Computer Using Agent, say the idea is to learn from how people use the tool. They acknowledge that the tool can make unwanted bookings or purchases, but add that a lot of work has gone into ensuring the tool asks before doing anything risky. “It will come back to me and ask for confirmation before taking any steps that may be irreversible,” Kumar said.
OpenAI also released a new “system map” today, detailing the issues that can arise with Operator. These include the possibility that commands are misunderstood or differ from what a user requests; to be abused by users; or are targeted by cybercriminals.
“It also poses incredible security challenges,” says Kumar. “Because your attack vector area and your risk vector area increase significantly.”
Operator will initially be available as a “research preview” for ChatGPT users with a Pro account, which costs a hefty $200 per month. The company says it plans to expand access and roll out the tool slowly, as it will inevitably make mistakes along the way.
In several demonstrations, Operator showed that AI has the potential to take on a more active role as a web helper. The tool features an external web browser and a chat window to communicate with a user.
At WIRED's request, 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 necessary information correctly to display the timetable, then asked for further instructions. If a user was logged into the Amtrak website or a browser profile with saved credit card information, Operator would be able to book a ticket, although it is designed to ask for permission first.
Kumar asked Operator to reserve a table at Beretta, a restaurant in San Francisco. The program went to the OpenTable website, found the right restaurant and looked up availability before asking what to do next. OpenAI says it's working with a number of popular sites, including OpenTable, to ensure Operator runs smoothly on them.
The new tool is based on OpenAI's GPT-4o AI model, which can sense a browser and web page and converse in typed text. The tool includes additional training designed to help understand how to complete tasks online. OpenAI will also make its Computer Use Agent available through its API.
