The browser became a workplace in January
OpenAI’s January 2025 Operator preview made a simple interface consequential: an agent could look at a webpage, type, click and scroll in a browser built for people. The important change was not that a model could produce text. It was that a model could operate inside the same pr
OpenAI’s January 2025 Operator preview made a simple interface consequential: an agent could look at a webpage, type, click and scroll in a browser built for people. The important change was not that a model could produce text. It was that a model could operate inside the same procedural spaces where people buy, book, submit and authenticate.
Operator was a research preview, and OpenAI described limitations and handoffs to the user when the system became stuck. That boundary matters. Browser control is not equivalent to dependable agency; it is a route from model output to external state.
The safety question therefore moves outward. Which pages may the agent visit? Which credentials may it use? Which clicks require approval? A browser agent needs logs and stop conditions as much as it needs visual understanding. The web was already an interface for humans. In 2025 it became an increasingly important interface for machines—and a new place to put authorization controls.
