This article looks at reports that Google Chrome quietly started downloading a 4GB on-device AI model called Gemini Nano to some users’ devices. That’s raised a lot of eyebrows about privacy and sustainability.
Gemini Nano runs locally on devices like smartphones and laptops. It’s supposed to help with stuff like scam-call detection, drafting messages, summarizing recordings, and analyzing Pixel screenshots.
Privacy advocate Alexander Hanff, also known as That Privacy Guy, noticed these sneaky installations. He pointed out that Chrome didn’t tell users or ask for permission.
Google claims Gemini Nano only installs on hardware that meets certain requirements. If your device doesn’t have enough resources, it’ll uninstall itself.
Since February, users can turn off and remove the model from Chrome settings. Still, nobody seems to know exactly how many devices got it or what this means for data protection rules.
What happened with Gemini Nano in Chrome
Gemini Nano works entirely on-device, so it cuts down on cloud processing for some tasks. That could mean better privacy and efficiency, but the way Google pushed it out—without asking—has people worried about control and transparency.
Google says the model will uninstall itself if your device isn’t up to the task. Since February, there’s an option to disable or remove it in Chrome’s settings.
Critics feel this kind of behind-the-scenes installation erodes trust. There’s also the question of whether shifting computing loads to your hardware, without clear disclosure, is really fair—or even allowed under EU rules.
For privacy advocates, the main issue is consent and knowing what’s happening. If a bunch of Chrome users got a huge AI model stored locally without a heads-up, that’s a pretty big deal in terms of data processing.
That Privacy Guy argues this undisclosed rollout could go against European data-protection expectations. It also muddies the line between browser features and hidden data handling.
Some people do see benefits in on-device AI, like better responsiveness and less cloud energy use. But only if users know about it and can actually make a choice.
How to check if Gemini Nano is on your device
Worried you might have this on your machine? You can check for the Gemini Nano files using your system’s file paths.
- Mac: Find an OptGuideOnDeviceModel folder with weights.bin inside ~/Library/Application Support/Google/Chrome/Default.
- Windows: Look for OptGuideOnDeviceModel under %LOCALAPPDATA%\Google\Chrome\User Data or C:\Users\[username]\AppData\Local\Google\Chrome\User Data.
If you spot those folders, you’re definitely not the only one. Many users have seen them after Chrome added opt-in and opt-out controls in response to privacy backlash.
Just finding the files doesn’t mean the AI model is running right now, but it does mean it was on your device at some point.
Managing on-device AI: options and controls
Google has added a way for users to manage on-device AI in Chrome’s settings. You can turn off the feature to block future downloads and updates, and you can remove the model if you want.
Windows users get a few extra steps to make sure everything’s gone after they close Chrome.
How to disable or remove Gemini Nano
If you want to turn off the feature or get rid of Gemini Nano, here’s what to do:
- Open Chrome Settings, head to System, and use the On-device AI toggle to turn it off.
- Quit Chrome, then check for the OptGuideOnDeviceModel folder and delete it if it’s there (see the Mac and Windows paths above).
- Windows users: You might also want to disable the Optimization Guide flag at chrome://flags. Set it to Disabled and restart Chrome.
Since February, Google says you can handle all this in Chrome settings, which should stop future downloads and updates. If you already had the model and your device can handle it, it’ll stick around until your resources drop and Chrome removes it automatically.
Privacy, data protection, and sustainability considerations
The bigger picture here? It’s all about privacy, data protection, and sustainability. The secretive rollout has sparked distrust among privacy advocates, who keep asking how much control people really have over software running on their own devices.
That Privacy Guy points out a real risk: if companies don’t disclose these deployments, they could violate EU GDPR rules. There are also growing requirements for transparency around data handling, energy consumption, and the full lifecycle of on-device tech.
For users and policymakers, a few things stand out:
- Find out if your devices have on-device AI, and check how it’s managed.
- Use any built-in settings to disable or remove local AI models when you want.
- Push platform providers to be clear about what’s installed, why it’s there, and how you can opt out.
- Look at regulatory advice on consent, disclosure, and environmental impact before using on-device AI features.
Here is the source article for this story: If You Use Google Chrome, Your Device May Have Secretly Downloaded a 4GB AI Model