What you need to know
A staggering report states Google installed an incredibly large AI model on users’ devices without their knowledge.
Word broke about Google reportedly installing a 4GB Gemini Nano model on users’ computers from That Privacy Guy. Their blog post goes into fine detail; however, one glaring key point is that this file was downloaded without users knowing. The post states that this model is there to carry processes, such as “Help me write,” AI-assisted browsing features, and AI scam detection while browsing.
Upon further inspection, Alexander Hanff (That Privacy Guy) says that this enormous 4GB download is triggered when “Chrome’s AI features are active, and those features are active by default in recent Chrome versions.”
Article continues below
This Gemini Nano model for Chrome’s alleged “benefit” has been an object of discussion for Windows users, who’ve reported a sharp decline in their computer’s storage (by 4GB to be exact). Moreover, users can’t exactly run from this unwanted installation. Hanff states users who’ve never interacted with Chrome’s AI features will still get this AI model downloaded onto their hard drives without their knowledge. Uninstalling it requires an extreme number of hoops to jump through, as Hanff states.
Users would need to discover the hidden path Google hid the AI model in on your machine, but it would just “silently re-download it on the next eligible window…”
Here’s where things escalate. Hanff states that, with Chrome 147 (which many should already have), there is an AI Mode button on the far right of your address bar. Users can open a new tab to see this appear. One might think that this on-device AI model is useful for that, to handle tasks you might require it to do. Hanff says otherwise. The post states, “every query the user types into it is sent over the network to Google’s servers for processing by Google’s hosted models.”
This development is creating a buzz, considering Google never asked users—any user—if they wanted to download this 4GB AI model onto their computer. On one hand, Hanff likens this behavior to what Anthropic did with Claude Desktop. Additionally, this might be illegal under U.K. and EEA law. This could be a breach of Article 5(3), with Hanff adding, “The 4 GB Gemini Nano weights file is information stored in the user’s terminal equipment. The user did not consent.”
We see these features, but did we need this?
Let’s get something out of the way: many of the features this AI model “helps” with are tools we’ve already been exposed to. Back in April, Google expanded its browsing assistant to more countries. The feature can pull content from webpages or videos you’ve visited to help you plan trips or summarize long articles online. What’s more, Nano Banana’s involvement lets you transform images you see into something entirely new.
This feature is here, even if it is a little concerning. Those AI scam detection parameters have been around for a while longer on Android and desktop. To walk the same line as my colleague, Jay Bonggolto, we usually have these “you’re in control” messages thrown at us about Google’s AI—most AI, too. But then, you see a report like this, and it begs the question: are we?
Android Central’s Take
I hate this. Flat out. I’ll give credit where credit is due, my colleague Nicholas Sutrich came up with this headline, and I couldn’t agree more. This is 4GB of my storage space, of my computer, that’s gone to… Google and it’s AI model? The answer is no, and the answer will always be no. I agree with Alexander Hanff, this is a breach in trust from a consumer’s perspective. No one was asked if they wanted this ginormous AI model on their computer. No one was asked if this was “okay to do.” It was just done, and Google is quiet about it.