Chrome Quietly Dropped Its On-Device AI Privacy Promise—Here’s What to Do

If you use Google Chrome and have noticed new AI features appearing in the browser, you may have also noticed that a key privacy assurance quietly disappeared. Until recently, Chrome’s documentation stated that its “on-device AI” features processed everything locally and did not send data to Google’s servers. That claim has now been removed, and there is evidence that Chrome is also installing a large AI model on users’ computers without clear, upfront consent. Here is what changed, why it matters, and what you can do about it.

What Happened

Multiple news outlets, including Decrypt and GIGAZINE, reported in early May 2026 that Google silently deleted a sentence from Chrome’s AI documentation that promised on-device AI would not send data to its servers. The removal appears to have happened without any public announcement or explanation. At the same time, Decrypt also reported that Chrome has been quietly downloading a roughly 4GB AI model onto users’ machines. Even if you delete this model, the browser may restore it the next time it updates, without asking for permission.

The exact nature of the data that might now be collected is not fully clear. Google has not issued an official statement about the change as of this writing. Some of the reports speculate that the on-device AI features—such as smart text selection, page summarization, or writing assistance—may now send telemetry or other data to Google as part of model improvement. But without confirmation from Google, the details remain uncertain.

Why This Matters

The original promise was important to privacy-conscious users. When a browser says a feature runs “on-device,” the reasonable expectation is that nothing leaves your computer. Removing that claim shifts the burden to users: you now have to assume some data may be transmitted unless you take action yourself.

The 4GB model installation adds another layer of concern. Unknown downloads that reappear after deletion are a poor practice, reminiscent of how some software installs unwanted components. Even if the intent is to improve browser features, the lack of transparency undermines trust. For users with limited disk space or data caps, a 4GB download is not trivial—and installing it without clear consent is a legitimate privacy and security issue.

What Readers Can Do

If you want to limit Chrome’s AI data collection and avoid unexpected downloads, here are practical steps:

  1. Review Chrome’s AI settings. Go to chrome://settings/ai and check what AI features are enabled. You can turn off experimental AI features and disable any that you do not actively use.
  2. Clear the downloaded AI model. Look for a folder called OnDeviceModel inside Chrome’s cache directory (location varies by operating system). Deleting it may remove the model, but be aware it might reappear after a browser update.
  3. Disable automatic downloads. In Chrome’s settings, disable “Help improve Chrome’s features and performance” under the privacy and security section. This may reduce unwanted background downloads.
  4. Switch to a privacy-focused browser. If you want to avoid these issues entirely, consider alternatives like Firefox, Brave, or Vivaldi. Firefox’s built-in AI features are optional and clearly documented, and Brave blocks many tracking scripts by default. Note that Edge and Opera also have integrated AI tools, so check their privacy policies if you switch.
  5. Use extensions to block tracking. Tools like uBlock Origin or Privacy Badger can help prevent unnecessary connections, though they cannot stop Chrome’s own internal telemetry entirely.

Sources

  • Decrypt: “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)
  • GIGAZINE: “Chrome removes claim that its ‘on-device AI’ does not send data to Google servers” (May 9, 2026)
  • Yahoo Tech coverage of the same story (May 7, 2026)

The situation is still developing, and more concrete information may emerge if Google responds. Until then, the safest approach is to assume that on-device AI in Chrome is not necessarily private and to take the steps above to regain control over your data.