Chrome has launched an on-device AI system that identifies and flags suspicious notifications on Android to protect users from scams—before they can cause harm.
The new system leverages Gemini Nano’s capabilities to analyze notification content locally without sending data to Google’s servers, maintaining user privacy while fighting notification-based fraud.
This automatic protection feature adds to Chrome’s existing security measures against notification abuse, and may expand to other platforms beyond Android in the future.
Chrome is rolling out a defense system that catches scammy notifications before they can do harm. The browser’s latest update for Android devices now uses on-device machine learning to identify and flag suspicious notifications that might be trying to trick you into downloading sketchy software, or handing over personal information.
Google confirmed Thursday that its new notification warning system is active, giving Android users an extra layer of security against an increasingly common form of mobile scam.
“When a notification is flagged by Chrome, you’ll see the name of the site sending the notification, a message warning that the contents of the notification are potentially deceptive or spammy, and the option to either unsubscribe from the site or see the flagged content,” the Chromium team said in a blog post announcing the feature.
Notification-based scams have become the digital equivalent of those annoying pop-up windows that plagued earlier versions of the internet—except now, they’re harder to spot. These new pests can even follow you even after you’ve left a website, which is why AI on your device works better than server-based AI, per Google’s assessment.
“Leveraging LLMs on-device allows us to see threats when users see them. We’ve found that the average malicious site exists for less than 10 minutes, so on-device protection allows us to detect and block attacks that haven’t been crawled before.” Google said in a separate blog post. “The on-device approach also empowers us to see threats the way users see them.”
Rather than sending your notification data to Google’s servers for analysis, Chrome leverages Gemini Nano’s multimodal capabilities to differentiate good from bad notifications. The machine learning model examines the text in each notification—including its title, body content, and any action buttons—then makes a determination about whether it looks fishy.
Image> Google Security Blog
This also means that Google does not have access to your notifications, because they remain encrypted and never leave your phone. Google claims it took extra precautions with the training data, using synthetic information generated by its Gemini large language model instead of actual user notifications. The company then validated this training data against real notifications that its security team gathered and had human experts classify.
The system adds to Chrome’s growing arsenal against notification abuse. The browser already revokes notification permissions from sites that Google Safe Browsing identifies as abusive, and offers one-tap unsubscribe for any notification on Android, whether it’s been flagged or not.
For now, the feature is Android-only since that’s where most notification spam happens, but Google hinted at possible expansion: “We will evaluate expanding to other platforms in the future,” the company noted.
The best part about this new protection is that it’s completely automatic, unlike other alternatives in which users typically have to deal with filters, exceptions, and other configurations, updating them as scams get more sophisticated. If Chrome spots something suspicious, then you’ll know about it before the scammers can make their move.
Some Programs require complex filters and configurations
With notification fraud growing more sophisticated, this machine learning approach offers a smarter way to spot the tricks that static blocklists might miss. Hopefully Apple will follow suit.
Generally Intelligent Newsletter
A weekly AI journey narrated by Gen, a generative AI model.
in well organized HTML format with all tags properly closed. Create appropriate headings and subheadings to organize the content. Ensure the rewritten content is approximately 1500 words. Do not include the title and images. please do not add any introductory text in start and any Note in the end explaining about what you have done or how you done it .i am directly publishing the output as article so please only give me rewritten content. At the end of the content, include a “Conclusion” section and a well-formatted “FAQs” section.
