Chrome quietly deleted its own privacy promise for on-device AI — here’s how to stop it
If you use Chrome and care about privacy, you might want to check what Google has quietly changed under the hood. Recent reporting from Decrypt and Yahoo Tech has revealed that Chrome removed a long-standing privacy promise that kept on-device AI from sending data to Google. At the same time, the browser has been silently downloading a 4 GB AI model that reappears even after you delete it. Here’s what happened, why it matters, and what you can do about it.
What happened
For years, Chrome stated in its privacy documentation that on-device AI features would not send any data to Google’s servers. That guarantee gave privacy-conscious users a reason to trust that their browsing data stayed local. But according to a Decrypt investigation, Google quietly excised that promise from Chrome’s privacy page. The new wording now says that on-device AI may still communicate with servers, opening the door for data collection that was previously ruled out.
On top of that change, Chrome has been automatically downloading a roughly 4 GB AI model onto users’ computers. This model is intended to power local AI features like smart text suggestions and page summarization. However, several users have reported that even if they manually delete the model, Chrome redownloads it on its own — sometimes within hours. The behavior suggests the download is not truly optional, despite what the settings may imply.
Why it matters
The shift matters for two reasons. First, it erodes a foundational privacy protection. When a company removes a published commitment without clear notice, it undermines trust. Users who rely on “on-device” labels to assume no data leaves their machine now have to question that assumption.
Second, the forced download of a large AI model eats up disk space and bandwidth without explicit consent. Many people may not even realize the model is there. For those on metered connections or with limited storage, this is a tangible inconvenience. More importantly, if the model can communicate with Google’s servers, it could transmit usage data, device information, or other signals that users did not agree to share.
What readers can do
You don’t have to accept these changes. Here are concrete steps to take control.
1. Disable on-device AI in Chrome’s settings.
Open Chrome and type chrome://settings/privacy into the address bar. Look for an option labeled “Allow on-device AI enhancements” or something similar (the exact name may vary by version). Toggle it off. This should prevent Chrome from using the AI model and reduce the chance of data transmission.
2. Delete the downloaded AI model — and prevent re-download.
If you already have the model, you can delete it manually. On Windows, check %LOCALAPPDATA%\Google\Chrome\User Data\Local State or similar directories. On macOS, look in ~/Library/Application Support/Google/Chrome/. However, simply deleting the file may not stop Chrome from downloading it again. Disabling the AI feature first (step 1) is essential to prevent that.
3. Consider switching to a privacy-first browser.
If Chrome’s direction concerns you, alternatives like Firefox, Brave, or even Microsoft Edge (with stricter controls) may offer better transparency. Firefox has a clear on-device processing policy and does not download large AI models without asking. Brave blocks many tracking features by default. Moving your daily browsing to one of these can give you more direct control over what runs on your machine.
Sources
- Decrypt, “Chrome Deleted Its Own Privacy Promise for Sneaky On-Device AI,” May 7, 2026.
- Yahoo Tech, “Chrome Deleted Its Own Privacy Promise for Sneaky On-Device AI,” May 7, 2026.
- Decrypt, “Chrome Is Quietly Installing a 4GB AI Model on Your Computer—And Putting It Back If You Delete It,” May 6, 2026.
Stay informed, check your settings, and don’t assume that “on-device” still means what it used to.