Google is taking a major step forward in online safety by launching a powerful suite of on-device AI security features across its platforms. The company announced on May 9 that it’s rolling out Google AI scam detection capabilities powered by Gemini Nano, its lightweight large language model (LLM), to combat rising online fraud on Chrome, Android, and Search.
The move will improve scam prevention in Chrome 137 for desktops, with Android versions following soon. Gemini Nano Chrome integration allows scam filtering to occur locally, without sending user data to the cloud. This on-device model ensures faster and more private detection of suspicious web activity.
Protecting Against Tech Support Scams and More
Currently, Google AI scam detection is focused on preventing remote tech support scams, schemes where users are misled into thinking their computers have issues. These scams typically lure victims into revealing sensitive information or installing malware.
To counter this, Gemini Nano evaluates sites using signals associated with fraudulent behavior, like use of the keyboard lock API, and flags them accordingly. These signals are passed to the Safe Browsing system, which can block deceptive pages before users are affected.
Chrome Security experts say the model runs efficiently on devices: “We limit the LLM’s activation, monitor GPU usage, and use asynchronous processes to avoid slowing down your browsing experience,” explained Google’s security team.
Spam Alerts, Notification Warnings Arrive on Android
In addition to website protection, Google is adding a notification warning feature on Chrome for Android. This on-device model analyzes incoming site notifications in real time, warning users of suspicious or spammy content that could lead to data theft or malware.
These upgrades follow Google’s recent integration of AI-powered protections in Android’s Messages app, where scam SMS filtering was added. The company is also developing an Advanced Protection mode in Android 16, echoing Apple’s approach by disabling JavaScript, 2G connections, and enabling anti-theft features like Theft Detection Lock.
Detecting Fraud Before It Reaches You
In 2024, Google’s AI models have helped reduce airline impersonation scams by 80% and cut down government service scams by 70%. Now, with Gemini Nano running locally, Google AI scam detection is positioned to become even more effective.
Google is also experimenting with new tools to detect voice-based scams, especially those involving attackers tricking users into opening banking apps during phone calls — a growing concern among cybersecurity experts.
About The Author
