Google Gemini Now Watches Your Chrome Screen in Real Time

Google is rolling out a feature in Chrome 149 that lets its Gemini AI see and select content from users' browser screens, framing it as a convenience upgrade. The company also announced that developers can now access computer use capabilities through Gemini 3.5 Flash, allowing AI agents to watch and act on screens. Critics warn this normalizes persistent screen surveillance, raising privacy concerns as the data flows to Google's servers and could be accessed by the government without a warrant.

Google is rolling out a feature that lets its Gemini AI see exactly what's on your Chrome browser screen — and the company is framing it as a convenience upgrade, not the surveillance on-ramp it actually is. The new "Select from screen" capability, arriving with Chrome 149, lets users highlight any text or image visible in their current browser tab and send it straight to Gemini. According to Digital Trends, it functions like a built-in screenshot tool — once activated, whatever you select gets automatically attached to a Gemini prompt. No manual description required. Just point the AI at your screen and ask. Here's why that should alarm every American who opens a bank statement, reads medical results, or communicates privately through a browser: Google is normalizing the idea that an AI assistant should have its eyes on your screen at all times. Digital Trends buried the implications entirely, calling the update a way to "remove friction" and move "beyond the traditional chatbot experience." The framing was pure product review — gee whiz, now the AI knows what you're looking at, isn't that handy. But the same day, Google announced something far more consequential. Developers can now access "computer use capabilities" directly through Gemini 3.5 Flash, which Digital Trends reported allows AI agents to "see, reason, and take actions across browsers, mobile apps, and desktop environments without relying on a separate model." Google says this targets "long-horizon tasks" like software testing and enterprise workflows. Translation: an AI that can watch your screen and act on what it sees is now available to third-party developers. The surveillance state has always wanted exactly this — persistent, real-time access to what Americans are viewing, typing, and reading on their devices. The Fourth Amendment requires a warrant for the government to peek at your screen. But when Google builds the architecture voluntarily and calls it an "assistant," there's no constitutional barrier at all. The data flows to Google's servers. A subpoena, a national security letter, or a quiet FISA court order, and it flows to the government. And nobody in Washington is asking a single question. No hearings. No letters. No proposed legislation. The same lawmakers who spent years grandstanding about TikTok's data access are silent while an American tech giant builds screen-level surveillance into the browser used by roughly 65% of the global market. The feature requires user activation — for now. But the trajectory is clear. Google is moving Gemini from a standalone chatbot to something that is "aware of what users are actively doing," as Digital Trends put it. Today you choose to share your screen. Tomorrow the AI suggests you share it. The day after, it's the default. The founders didn't draft the Fourth Amendment so Americans could voluntarily surrender their privacy to a corporate middleman that hands it over on request. The question isn't whether Google will abuse this. It's whether the government even needs to ask anymore.