Google Security Blog

Chrome 137 introduces on-device AI-powered scam detection using Gemini Nano LLM to identify tech support scams in real time. When a page triggers suspicious signals (e.g., keyboard lock API usage), Chrome runs the LLM locally to extract security signals and sends them to Google Safe Browsing for a final verdict. The on-device approach enables detection of short-lived malicious sites (average lifespan under 10 minutes) and sites that render differently to evade crawlers. The feature is available to Enhanced Protection users, with Standard Protection users benefiting indirectly via updated blocklists. Future plans include expanding detection to package tracking and toll scams, Android support, and defenses against prompt injection attacks.

Google Online Security Blog: Using AI to stop tech support scams in Chrome