- Google Chrome silently installs a 4 GB AI model on your device without consent - https://news.ycombinator.com/item?id=48019219 - May 2026 (1143 comments)
- we're gonna keep posting this until morale improves!!! google bad!!! /s
- Ever since IO earlier this year when google showed their AI strategy I switched to firefox and duckduckgo and couldn't be happier with the decision. I am by no means anti-AI but users should be able to choose when they want to use these tools, google seems to want to shove it down everyone's throat.
- Did a similar thing, but went with Kagi + Orion. Also happy to have a few more free cycles from not thinking about Google.
- IMO If you're into tech and still use Chrome*, that's on you. If you are not, you probably don't really care unless you need extra space on your PC.
*Except in your job, since you probably obligated to use.
- Switched from firefox few months ago, I don't like google, but firefox has only few percent of market share currenly, many pages simply do not work properly with it, plus it has bugs on macos (like onmousover stuff), which simply make it unusable. Safari is fine, but also many websites (which sometimes you need to use like banking or gov) are not tested on it. The overall browsers situation is less than ideal.
- > ...many pages simply do not work properly with it,...
Using Google's websites much, are we? FF works great everywhere except there. I think the official term for the bugs on these pages is "oops"...
I use FF everywhere except on G pages, where I use Brave.
- That is weird, I've daily driven firefox for the better part of a decade (aside from when my employers have required chrome) and seldom encounter issues. I'm curious where you're hitting these. In fact, since ublock origin got removed from Chrome, the experience is far better on firefox.
Also, if everyone chooses to not use firefox because it has low market share, it'll remain low market share forever.
- > many pages simply do not work properly with it, plus it has bugs on macos (like onmousover stuff),
I use firefox on mac and I have simply no clue what you're talking about. Tons of people use firefox on mac quite successfully...
The only page that I know of that doesn't work is google earth - it doesn't work on linux either. (Technically it does, just incredibly laggy compared to chrome)
I can't think of a single macos specific bug. Or a single mouse over related bug.
- I think it's a stretch to call firefox on mac "unusable". Like once a month I'll have something not work, and half the time it also doesn't work in chromium (both are the developer's fault). And when it does work in chromium, browsers are free so just switch over there and switch back. And to top it off, there are endless free flavors of chromium like brave etc, that virtually never have compatibility issues without needing to go full google.
- I’ve been a daily Firefox user on macOS and windows for a decade. I can count on one hand the number of sites that don’t work.
What sites do you have in mind?
- I have been using Firefox for years (decades) and I rarely encounter a page or site that doesn't work... a handful perhaps, but it's a complete non-issue for me. If anything, my ad-blocker can sometimes cause issues and I'll disable that and everything is good again.
- Technically speaking, Safari is the second most used browser. It’s quite surprising how often I get these “unsupported browser” popups. I still use it as my primary and have FF as a backup. No Chrome needed.
- Don't be so dramatic. Firefox on macOS is perfectly usable.
- daily drive Zen on macos with zero issues and quality adblocking support.
- Really? I use Firefox as my personal browser and everything works fine, including Google sites. Very rarely there’s a government site that needs Chrome, but I definitely wouldn’t say it’s “many” sites.
- Any page can silently trigger an additional multi-gigabyte download for Chrome users by just calling this API:
Since the model is installed once per browser, LanguageModel.availability() could probably also be used for fingerprinting.await LanguageModel.create()- Indeed, you could even do something like this:
It even works on non-Chrome browsers.await fetch("/multi-gigabyte-download");- i think the point is that one actually burns disk space, while the other is just a nuisance for people on plans with limited data
- You can store up to 80% of free disk space with indexeddb
- using a browser is an easy way to burn disk space and limited data
light browsing is not a thing unless you somehow ensure javascript is disabled and all downloads are blocked
- Just the availability wouldn't be that bad from a fingerprinting standpoint (getting one bit that a majority of Chrome users have is just the same bits you already have, usually), except, it also exposes whether the underlying hardware is "eligible," and once it's running, you can also benchmark the language model performance. It's a mess. I think it might also be broken and work in iframes, which would be an even bigger mess; there are a few bug reports suggesting this although many of them look like slop.
This feature was massively bungled; I actually don't overall hate the idea of it (having a shared, pre-downloaded model that can run effectively from JS is kind of awesome versus sites downloading stuff into LocalStorage to use with hacked up wasm/webgl inference engines), but it really, really needed a permissions dialog and a proper anti-fingerprinting model.
- > really needed a permissions dialog
Why?
- Users with data caps are an obvious reason, for example, see this frustrated user report here: https://support.google.com/chrome/thread/415181794/chrome-al... .
To your point about `fetch` in the sibling comment; yes, I would also be annoyed by a website which downloaded 4GB in the background without asking me, too.
I don't think this is some moral outrage, to be clear; it's well within Google's rights, it doesn't seem "sketchy" at all, and it is kind of a cool feature, but it feels like they could have done a lot better by just making it opt-in and a little less fingerprintable.
- In their defense, it's an astonishingly good model for this size and you can use it for all sorts of cool stuff.
Little demo of using this local model to inject AI into a page with a monkeyscript: https://www.youtube.com/watch?v=MPi33D8DoQ0&t=3000s
- yeah I'm lowkey pissed off people keep insulting google for giving them an offline version of coding enabled wikipedia that they get literally 0 data analytics from (just turn your wifi and ethernet off and use it offline after downloading it)
what part of the 4Gb file offended people?
the fact that ai runs on the edge reliably now?
- Memory prices will adjust accordingly as this becomes the norm.
- > yeah I'm lowkey pissed off people keep insulting google
This is the kind of bootlicker that will usher in the new distopia.
Thanks
- you're welcome for the total lack of reading comprehension on your part.
the edgy kids all hate google we get it but how the fk is a free offline ai model even a bad thing for privacy or even benefitting google in any way other than giving away privacy preserving tech for free
they obviously aren't good intentioned here, and benefit from the aggregate of data but OpenAI and Anthropic might as well be enemies of all humanity given they contribute nothing back to the knowledge that is useable offline without their exorbitant api costs
google should be better about their more evil tendencies but they now have sufficient competition that bootlicking is not an important thing to point out in the fight over open source(ish) models that allow people more freedom not less
its all moot when nobody can afford hardware in the first place that can run ai models so I guess I'm just yelling into the void here
- That does nothing to defend Google since the quality of the model is irrelevant to the accusation.
- You'd think that if a technology is so good you don't need to force it on people.
- It certainly shouldn't require masses of blank eyed fans all making highly emotional pleads to protect the sensitivity of their favorite logo.
That's gross even by Microsoft standards.
- blank eyed critics not understanding what fight they are fighting are funnier yet
muhhhh logo bad thoooo
- I personally use Firefox mainly due to extension support so I don't experience this. However, as a developer, the embedded Gemma local model in Chrome is vastly better to be included as a one time download than it is to download an AI model for every site you visit, as AI and especially local AI becomes increasingly more common. It is the same idea that Apple and Android and Windows have with their built-in foundation models in the OS. So in that case I can appreciate what Chrome is doing, as it will be saving a lot of bandwidth over time.
People have asked me what use cases this has, and I've been making little apps with on-device AI such as a calorie estimator and tracker for food, or a page summarizer or translated (the latter of which Firefox actually already has using a local model). Why pay for a cloud model when the user has a perfectly good model themselves?
- To prevent Chrome from automatically downloading the model, go to Settings → On-device AI.
- I checked, and I do have the 4GB file on my system.
The “Local AI” toggle is enabled. Chrome never asked for my consent to enable it.
Screenshot: https://imgur.com/a/S4WTHxM
The 4GB file is located in C:\Users\[user]\AppData\Local\Google\Chrome\User Data\OptGuideOnDeviceModel\2025.8.8.1141
Screenshot: https://imgur.com/a/wvTqfQM
- i love the built in AI - i created this tic tac toe game with an AI trash talking opponent -> https://taf2.github.io/ai-tic-tac-toe-trash-talk/
- The page uses 10% of my CPU while doing nothing on it.
- At least no regrets purchasing that expensive CPU and the solar panels
- wait till you open my website, it will crash your GPU just from rendering the fractal
- I'm a big fan of local models, and moving inference from the cloud to local machines is great, but there's a couple potential problems with this. LLMs take significant (V)RAM resources to run (which is in short supply on consumer hardware), and we don't know that Google won't send local conversation logs to their own servers (kind of surprised if they don't). So if you think you're having a private local conversation in Chrome, I wouldn't be so sure.
- From a year ago. They talked about adding the ability for Chrome to install Gemini Nano on-demand when a web app wants to use it. I guess this is the next step, making it always available by default.
Practical built-in AI with Gemini Nano in Chrome
- This looks like AI-written blather, but it links to an article with useful information by someone who actually checked:
https://tech.yahoo.com/ai/gemini/articles/fact-check-google-...
- If you had bothered reading the article you would realize you only provided redundant information.
- Thanks! Very informative. Lot of new details including the answer from Google and several experiments
- Only problem is that it doesn’t prompt “do you want to download this/enable AI”, right? Otherwise, local AI is exactly what most people want instead of needing to pay rent to big tech companies for AI use*. Open-weight models would also be nice, but I imagine the people who care about that _and_ want to use Chrome are using Chromium or de-googled Chromium.
* other than how everyone is already paying rent in the form of increased energy costs due to increased demand from this massive ramp of data center development everywhere, regardless of if they want to use AI or not
- I'm sorry but the AI model that was forced to write this article is struggling to explain why this is a problem. I get that a chrome update that suddenly balloons to 4gb+ is stupid, that's fair, but I'm not sure I understand the rest of the issues. They don't like the off-device AI features Google is forcing into everything, but they also don't like the on-device AI features since they don't do enough?
Aside from taking up a lot of space as a web browser, I'm not sure I get it. Their explanation that Apple's version of this is fine but Google's isn't is wild too.
- The reality is that the move is great, it's a very cool and nice stuff, just that it takes bandwidth and space without letting the user know.
A solution:
New AI features are available for use offline, they will make you able to translate offline, get answers, summaries, etc, would you like to download / install them (~4 GB)?
It is going to fix the experience in the UI. It's a significant misstep by Google in its form (probably lazy/hasted/bureaucratic release), but on the move itself, this is actually a very user-friendly initiative from Google.
It's quite unfair in that specific case to say that Google = evil, and when Qwen = good. It's just about informing the user better so the bandwidth and space is not wasted. Giving user choice.
They will fix it eventually, especially after raising the issue.
But shouting here "Google = bad, me uninstall Google, if you use Google Chrome you are an idiot" are not a productive feedback for a product owner there.
- > But saying "Google = bad, me uninstall Google, if you use Google Chrome you are an idiot" are not a productive feedback for a product owner there.
I wasn't? I hope that wasn't directed at me since we seem to kind of agree otherwise.
Unfortunately the vast majority of Chrome users wont understand that prompt. Google decided that the Chrome installer should be 4GB, that seems like a large web browser, but is it that crazy compared to anything else? Teams on my Mac is taking 1GB+ just in Applications (not to say I'm terribly happy about that). Chrome installs its own updates and has for a very long time, ensuring we dont have to support super old browser versions, I'm not sure where the line is.
- I removed the model and then removed Chrome too. Good effing riddance.
- You can try the model here: https://chromeai.org/ -> Press the left edit button to start a new context.
It's free, multi-language, well tested, respecting privacy (no cloud needed), nothing extra to install. Quite nice actually.
Like what iPhones do, and everybody is ok with it. Chrome is not just a browser, it's a Window (ahem) to the web, almost an operating system considering the wide scope.
Average game is 80 GB, Call of Duty 200 GB+, etc
It's a quite oriented title:
Google Chrome silently installs a 4 GB AI model on your device without consent
The same news can be read as:
Amazing, Chrome now includes a fully offline AI so you don't have to send your secrets to ChatGPT
A proper journalist would have found the middle ground, explained that this is by default, can be made optional and to raise the issue to Google as "Cons: this uses bandwidth and user is not aware of it".
- Story from May;
Discussions: https://news.ycombinator.com/item?id=48019219
- So, Did Google respond?
- Google’s PR team is so bad at managing their reputation that their response wouldn’t have mattered. Apple can download a 4GB Apple Intelligence model onto your phone (which likely has lower storage than your PC) without any controversy; Google cannot.
- In the only way they know how :-)
- The amount of Google bootlickers and sycophants in this thread is genuinely concerning.
