I can’t see any issues here.
You agree to download and install Chrome and all its elements.
The size of the file is therefore irrelevant.
Obviously it uses the cloud, you can’t run an AI model locally and it would be impossible to secure anyway.
All solved with one simple step, DON’T INSTALL THIS GARBAGE!
Unfortunately it is not so simple. Microsoft Windows uses Copilot inside the system. Google Chrome uses Gemini models. Even the duckduckgo uses an optional AI. Nowadays any computer touches the internet touches AI systems directly or indirectly.
- Use as much as possible the local software instead of cloud-based softwares
- F*** the big tech companies. Use Linux, use librewolf or brave or any suitable browser instead of Chrome and Mozilla.
- Use firewall all over your system.
- Use DNS filtering
- Use custom roms In your android
- Fuck Google search. Use another search engine.
- Use Neovim or VSCodium instead of motherfucker VSCode
- Use self hosted cloud services
You cannot avoid 100%, but you can decrease the collateral damage in this way.
Not simple but worth the effort IMO. Totally agree with your list, use FOSS when possible, Adguard Home makes for great DNS filtering, I’ve been a fan of Kagi for search, gotta pay a little but get a ton of control over your search engine.
If someone is comfortable with computers I highly recommend looking into a VPS, you can get one cheaper then cloud services and use tools like Yunohost to make getting started self hosting not hard.
Man I sure do wish Linux mobile was a thing. I really want a full on Linux phone.
There are some early-days options for such things.
true
Same.
Yeah but that hardware man.
I think some action by EU is in order - something like “u have to release source code” and allow boot unlock - maybe with some conditions
How unsurprising anymore in this hellish world where corporates hate your desire for anonymity… but try to hide theirs, such as dark expense accounts, tax evasion, secret offshore banking accounts, connections with crime and hate groups, etc.
The increasing enshittification of every service pushed me to GrapheneOS long before Google could force this shit on me
The article doesn’t really say but this is just for desktop chrome right now right? I’ve long had chrome disabled since graphene isn’t an option unless I build it myself but I do worry about that pesky web view that snuck it’s way into everything.
Desktop chrome only
Huh? What’s the name of it then? My Chrome is but an .exe with 4mb to it and I don’t see any 4GB…
Edit: From Article, “Two weeks ago I wrote about Anthropic silently registering a Native Messaging bridge in seven Chromium-based browsers on every machine where Claude Desktop was installed [1]”
So, morons who installed shitty AI are getting AI shafted - makes ALL the sense to me.
It’s stated in the second paragraph:
This week I discovered the same pattern, executed by Google. Google Chrome is reaching into users’ machines and writing a 4 GB on-device AI model file to disk without asking. The file is named weights.bin. It lives in OptGuideOnDeviceModel. It is the weights for Gemini Nano, Google’s on-device LLM. Chrome did not ask. Chrome does not surface it. If the user deletes it, Chrome re-downloads it.
Its a separate file (as any 4GB download would be).
It’s funny because they’re trying to find ways to cut cloud costs by offloading to users, but when that’s not a concern, they shove everything into the cloud and then ensure no local running option is available or viable.
They want all of your data in the cloud so they own it.
They want all their crap on your device, so you pay for it.
deleted by creator
This is very alarming. My eyes have never been opened so widely as they are in the last two months since I started ungoogling and FOSSing. This post has veritably split my eyelids.
Edit: Since reading this thread I have installed Shizuku + Canta and removed Chr9me and about 50 other pieces of bloatware from my phone.
Props to @zerozaku for the suggestion.
uninstalled Chrome a looooong time ago on my Win 10 machine
Hope they didnt hardcode it!
They did. If you remove it, Chrome detects that the folder is missing and it downloads it again. Just don’t use Chrome.
Interesting. This update is not on my phone or my PC, which are both up to date. Wonder who it’s installing on.
this makes zero sense because it’s on device, it’s no difference than the damage that just owning a phone is costing, are people here special?
I swear people just want a reason to freak out. Atleast make sense if you’re going to post such a stupid article title.
So we now have a four-way evidence chain - macOS kernel filesystem events, Chrome’s own per-profile state, Chrome’s runtime feature flags, and Google’s component-updater logs - all four agreeing on the same conduct, and the conduct is: a 4 GB AI model arrived on this user’s disk without consent, without notice, on a profile that received zero human input, in a window of 14 minutes and 28 seconds, on a Tuesday afternoon.
How do we uninstall or block the download?
Uninstall chrome
Can you even uninstall chrome on an android phone? I only get the option to disable.
Use Shizuku and Canta to uninstall any uninstallable app. Or if you don’t want to bother, just disabling works fine too as long as you are not worried about the storage.
Thank you for the excellent suggestion. Worked perfectly. Managed to uninstall about 50 pieces of bloatware from my phone, starting with Chrome.
Props to you @zerozaku
Happy to help! Real props to the devs of these amazing apps.
Hear hear. 📢
Learned about this the other day and gave it a whirl, worked great, felt reminiscent of old school iPod jailbreaking shenanigans, but I had no issues. Easier (in a way) than adb!
You can use adb on a computer to remove chrome.
this is very helpful info, thank you, didn’t realize this was possible.
That depends on the ROM you are using.
The one i am using (https://iode.tech/) is using a firefox based browser that you can actually uninstall.
Probably not stock Android. I’m on GrapheneOS and it doesn’t come with Chrome at all. But I don’t think the article is claiming it happens on Android.
And install Firefox or one of its many forks.
So it just to the Chrome app?
The article actually gives 3 options:
The only ways to make the deletion stick are to disable Chrome’s AI features through chrome://flags or enterprise policy tooling that home users do not generally have, or to uninstall Chrome entirely
- It can probably be reverted at their whim at any time
- You probably don’t have access to it
- It is the most realistic option, just use another non chromium browser
Even Chromium should be fine. I doubt it has the branded Google AI features.
More difficult to remove than install. Adding the file took zero clicks. Removing it requires (a) discovering the file exists, (b) understanding what it is, © navigating into a hidden user profile path, (d) deleting it (and on Windows, also clearing the read-only attribute first), and (e) accepting that Chrome will silently re-download it on next eligible window unless the user also navigates chrome://flags, enterprise policy, or platform-specific configuration tooling to disable the underlying Chrome AI feature [5]. None of those steps is documented in the place a normal user looks - none of them is even hinted at in default Chrome.
This is 5: https://pureinfotech.com/stop-chrome-gemini-nano-download-windows-11/
Obviously only windows focused, so how other platforms stop would require more searching.
I don’t have Windows 11. Still on 10 until October then switching to Linux.
Can someone ELI5 why they are doing this? I thought all the AI shit was in the cloud?
The cloud being a bunch of computational power (servers). A bunch of phones in a network also can be utilized for said computational power. Passing the savings on to you! ;)
They need their features to work offline too probably.
It’s also cheaper, if they can offload a portion to the user’s computer.
Cheaper for them, that is.
What I want to see is throttleable models, kind of like progressive JPEG, where the default model is “nano” and it has a watch function that analyzes if more tokens might be needed for a certain task and scales up as needed — identifying if the resources are too much for the device and offloading to the cloud (with explicit permission) only if (but always if) needed. Over time as the technology improves, larger models move to the endpoint.
And then people could have a basic set of sliders: on-device only, on-cloud only, or somewhere in between, based on the user’s preferences.
That’s basically model routing, and has existed a while. Open AI’s GPT-5 and llama-swap do that, for example. If the task is simple, it uses a smaller, less intensive model, and only uses the slower, larger one of the task is more complex.
Though most tend to operate with models on the same device/service, rather than a model run elsewhere.
In an internet browser?
Yeah, even there. A page loading is one thing, but browser features are somewhat independent of the content. There’s also a good chance this is being used as a hook for other Google products like Drive or Docs (which are basically websites under the hood) to allow offline file management, creation, etc.
It’s a bad choice, but it wouldn’t be the first bad choice Google has made.
Well everything else is in it.
Shit, Chrome supports the use of COM ports. It’s an OS within an OS.
It’s like the new Bitcoin miner.
The Gemini on my device just became 10x smarter. Google Assistant didn’t know ‘what was the temperature a year ago’. and Gemini has no problem with it.
Assistant was made 100x stupider over the course of the years. I clearly remember in 2020 asking basic questions like “where’s Lima” and it replied “Peru”, while since 2023 it replied “ok, starting navigation to Lima, Peru, estimated time 2 days and 10 hours”.
Around 2023 also they made the voice assistant completely useless by changing the responses from “for this answer open the link in your notification bar on the phone” to “sorry, I didn’t understand”, which was pissing me off too much. I completely stopped to use it as 9 questions out of 10 would be “sorry I didn’t understand” instead of “I understood but I’m not programmed to give a voice answer, use your phone”.
Navigation by voice was a complete disaster.
“Navigate to <CONTACT NAME>” - “Ok navigating to <a random business 1000 km away with a name slightly similar to what I said>”
Or, it dropped the road name: you asked to go to “street name, city” and it placed the destination on the geographic center of that city
Also the navigation instructions would have maximum priority and would play even when listening to a command.
“Send a message: I’m going to be late <assistant overlaps my speech>” - “Sending: I’m going to be late on the next exit take the right lane then after 600 meters turn left. Ready to send it?”
“Call Anna” - “OK, I call Daniel”
I’m not sure if Gemini is that much better or they slowly degraded Assistant. I distinctly remember being able to ask certain things of Assistant and a year ago it stopped working. Assistant was still there but many requests no longer worked.
Android Auto was recently upgraded to Gemini. On a drive today, I asked Gemini to explain the privacy implications of bill C-22 (Canada), and it gave me very detailed clear explanation of the pros and cons of the bill, asked if I was a company or person and then gave me highlights relative to who I was. Pretty impressive actually.
