Google now wants its AI to run your computer for you — and it wants you to hand over the keys willingly. The company rolled out "computer use" for its Gemini app on macOS this week, giving the AI agent direct access to click through your apps, type in your forms, rifle through your local files, and browse the web on your behalf. The stake for ordinary Americans couldn't be clearer: the same company that already hoovers up your search history, email, and location data now wants an agent sitting on your desktop with the power to act as you.
Google frames this as convenience. According to The Verge, the company suggests it might be useful for "speeding up workflows, filling in forms, making presentations, and listing items for sale." TechCrunch, which framed the rollout as Google being "a little late to the game" behind OpenAI, Anthropic, and Meta, echoed the same pitch: tedious tasks, batch-editing photos, organizing files. TNW likewise ran through the same examples — pulling data from PDFs, turning dashboards into slides. All three outlets led with the productivity sales job. Buried in their coverage was the part that matters.
TNW reported what the other two soft-pedaled: Google warns that Gemini can make mistakes, including permanently deleting files or sharing them with others. Read that again. An AI that can delete your files or send them to someone else — and Google's answer is a permission prompt and a 24-hour backup window. After that, backups are gone. The Verge, notably, didn't mention this risk at all. TechCrunch skipped it too. Only TNW put it on the page.
Then there's the Apple angle. The Verge reported that Apple is preparing new restrictions for apps requesting "full disk access" on Mac, specifically citing "potential risks from AI agents with computer use." A preview landed in the macOS 27.3 beta this week. In other words, Apple sees the danger — Google is just moving faster than the guardrails. TechCrunch and TNW didn't touch this.
Users do get some controls. You pick which apps Gemini can access. You can pause it. You must approve key actions. It runs in a background window with a live preview so you can watch it work. But let's be honest about what this is: a panopticon dressed up as a personal assistant. Every permission you grant, every folder you connect, every form it fills — that's behavioral data flowing back to Google. The feature requires a paid subscription — AI Plus, Pro, or Ultra — so you're footing the bill for the infrastructure that watches you. It's only available in the U.S., not Europe or the UK, where privacy law still means something.
The pattern is unmistakable. Big Tech builds surveillance infrastructure one permission prompt at a time. Each feature is "helpful." Each grant of access is "optional." Each risk is disclosed in fine print nobody reads. And before long, the AI agent on your desktop isn't a novelty — it's the default, and opting out means falling behind.
Google says you're in control. The question is how long that control lasts once the feature becomes expected, the permissions become routine, and the competitors all demand the same access.







