Just because you can, doesn’t mean you should. That’s the message from the latest Chrome upgrade to surface in pre-release versions of the world’s most popular browser. Google is upgrading everything with AI — as just seen with Gmail and Personalized Gemini AI. Now Chrome is the latest platform giving users a major decision.
While deploying on-of off-device AI to write emails or search through files carries risk, its use to detect malware and scams is less controversial. Especially when it’s done on-device, without any of your personal data being sent to Google’s cloud.
That’s why this AI-fueled scam detection was so welcome when it was first detailed in late 2024. Of all the AI use cases, this real-time protection against fake websites, dangerous installs and downloads, or scam texts and emails is much needed.
But Google is now opt-out led. There has plenty of much controversy around default or hidden AI processing. Especially whenever there’s any implication that any AI models are being trained on user data or there is any form of data harvesting or human review.
There two competing imperatives have now collided. Per @Leopeva64 on X, the security researcher who found the AI upgrade in the first place, “Google has now added a new toggle that lets you delete the GenAI models (which power this feature) from your device. Turning it off also disables the feature itself.”
The pre-release “On-device GenAl Enabled” toggle, Google says, “powers features like scam detection locally. Turning this off deletes GenAl models from your device and disables these features.” Very much all or nothing, then. At least for now.
One thing worth point out is the word “like.” While scam detection is non-contentious, on-device AI can be turned to a wide range of other purposes that are less binary in their widespread usefulness. We don’t know what they might be, but we can speculate based on what we’re seeing elsewhere. Advertising, commerce, personalization… maybe.
Enhanced Protection itself is now new to Chrome. What’s new in the addition of AI. It’s not yet clear what safe browsing optionality will be available and how it compares in its effectiveness once all this goes live. On-device AI applied purely for security and user protection, is categorically a good thing. Ideally that can be toggled on its own.