https://blog.google/technology/safety-security/how-were-using-ai-to-combat-the-latest-scams
Google has unveiled a new suite of artificial intelligence-powered security features designed to protect users from evolving online scams across its Chrome browser, Search platform, and Android devices. These enhancements leverage on-device machine learning to detect and block fraudulent content more effectively.
Central to this security upgrade is the integration of Gemini Nano, Google’s on-device large language model (LLM), into Chrome’s Safe Browsing feature starting with version 137 on desktop computers. This implementation allows for real-time analysis of websites to identify potential scams, even those that haven’t been previously cataloged in threat databases.
“The on-device approach provides instant insight on risky websites and allows us to offer protection, even against scams that haven’t been seen before,” Google stated in its announcement. “Gemini Nano’s LLM is perfect for this use because of its ability to distill the varied, complex nature of websites, helping us adapt to new scam tactics more quickly.”
The technology is already being deployed to combat remote tech support scams, which typically attempt to trick users into revealing personal or financial information by fabricating computer problems. When users visit a website, the LLM analyzes the page for suspicious elements, such as the use of the keyboard lock API, which is commonly employed in tech support scams. These security signals are then processed by Safe Browsing to determine if the site is fraudulent.
Google Chrome Security engineers explained that they’ve designed the system to be resource-efficient: “We carefully manage resource consumption by considering the number of tokens used, running the process asynchronously to avoid interrupting browser activity, and implementing throttling and quota enforcement mechanisms to limit GPU usage.”
Google plans to expand this capability to detect additional scam types, including package tracking and unpaid toll frauds, with Android support for Chrome scheduled for later this year.