Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Google Gemini you have a model You may soon be able to act like an IT admin taking charge of your faulty work laptop: moving the cursor, clicking on things, and typing in forms, all while you watch.
The company announced the launch on Wednesday Gemini 2.5 Use the computer The model is in preview in its Application Programming Interface (API), allowing developers to create agents that can interact directly with the computer’s user interface. This visual understanding of the screen and its elements, through taking screenshots and analyzing them, allows an AI agent to do anything on a computer that you might be able to do, allowing you to delegate all kinds of tasks to an automated tool.
Don’t miss any of our unbiased technical content and lab reviews. Add CNET As Google’s preferred source.
Although it mostly works in web browsers at the moment, it can do some actions on the mobile interface as well. Google said the model, based on Gemini 2.5 Pro, has not yet been optimized for desktop OS-level control.
This new model is part of a growing trend towards… Amnesty International agenta technology that allows the model to bypass the chatbot box and take actions in the real world of a computer interface. Tools like ChatGPT Proxy It can actually do things like I order you a pizzaalbeit with some restrictions. Some agent tools may replace mundane tasks in the workplace or in customer service interactions. At the same time, AI companies are increasingly bringing more and more technology to bear Things we normally do on our own In the chatbot interface. With agents, the conversational nature of a tool like Gemini replaces the point-and-click you’re used to.
Watch this: New poll shows growing use of AI among kids, controversy over Xbox Game Pass pricing, and California law promising to cut advertising | Technology today
As AI agents gain the ability to manipulate websites and potentially applications on the computer itself, AI makers will need to take significant security precautions. Google acknowledged as much in the blog post announcing the model, saying: “AI agents controlling computers present unique risks, including intentional misuse by users, unpredictable model behavior, instantaneous injection and fraud into the web environment.”
Google said it trained the model specifically to address those risks, including the ability to recognize when it is given a “high-risk” command such as sending an email or buying something. The form may also require the user to confirm before taking a high-risk action.