Google Deploys On-Device AI Security Measures to Identify Scams in Chrome and Android Platforms

Google has announced the rollout of new artificial intelligence (AI)-powered measures designed to combat scams across its platforms, specifically Chrome, Search, and Android. The company will utilize Gemini Nano, its on-device large language model (LLM), to enhance Safe Browsing in the latest version of Chrome on desktop systems.

This innovative on-device approach enables immediate identification of risky websites, offering protection against scams, including those that are newly emerging. The adaptability of Gemini Nano’s LLM allows for rapid responses to evolving scam tactics by distilling the complex nature of various websites.

Google has previously implemented this AI-driven strategy against remote tech support scams, which typically attempt to deceive users into disclosing personal or financial information under the guise of resolving fictitious computer issues. The LLM evaluates web pages for specific indicators characteristic of such scams, including the use of the keyboard lock API. The signals identified are then relayed to Safe Browsing to assess the likelihood of the webpage being a scam.

Google’s Chrome Security team has emphasized the importance of efficiency; the LLM is designed to activate only when necessary and runs locally on the device. The management of resource consumption is prioritized through careful monitoring of token usage, asynchronous processing, and the implementation of throttling and quota enforcement to minimize GPU utilization.

Looking ahead, Google plans to extend this AI functionality to detect additional scam types, including those related to package tracking and unpaid tolls, with future updates expected for Chrome on Android.

The company also disclosed enhancements to its AI-based scam detection systems, resulting in the identification of 20 times more deceptive pages and blocking them from search results. This initiative has led to a significant reduction in schemes impersonating airline customer service representatives by over 80% and those mimicking official resources like visas and government services by more than 70% in the current year.

A new warning feature is also being introduced for Chrome on Android. This feature employs an on-device machine learning model to alert users about unwanted notifications from malicious sites, which aim to trick users into downloading dubious software or divulging sensitive information. When such notifications are flagged, users will see the site’s name, a warning about the potentially deceptive content, and options to unsubscribe or review the flagged material.

These updates follow Google’s previous deployment of AI-powered scam detection capabilities in the Messages app for Android, and similar measures introduced for identifying scam calls last year.

Moreover, as Google prepares to launch the Advanced Protection feature in Android 16, this initiative incorporates robust security measures including the disabling of JavaScript, activation of default security features like Theft Detection Lock and Offline Device Lock, alongside enhanced spam protection in Messages.

In related advancements, Google is also reportedly working on features that can detect scams orchestrated during phone calls, which attempt to manipulate victims into accessing their banking applications.