GOOGLE has announced a broad upgrade to Android scam protections, leveraging Gemini AI to tackle increasingly personalised telecom fraud and social engineering. The update brings real-time scam warnings into the notification shade for a wide range of Android devices, expanding beyond Pixel to phones from major brands.
Google Messages gains a hybrid defence that uses Private AI Compute in the cloud together with on‑device Gemini Nano to analyse conversations and scan malicious links in advance, while masking screen sharing and blocking bank‑call hijacking attempts. The system is described as capable of recognising classic trickery, such as scammers starting with a wrong-number chat to gain trust, before escalating to personal information requests.
The architecture consists of a two‑layer design: Private AI Compute handles the most advanced Gemini models for multi‑turn conversations in a secure, hardware‑encrypted cloud space, ensuring messages remain inaccessible to third parties; Gemini Nano runs offline on the device to check unknown links before the user taps them. The rollout starts with Google Pixel 10 series and newer, and extends to flagship devices with new-generation NPUs, including the Samsung Galaxy S26 series, Galaxy Z Flip8 and Galaxy Fold 8.
Beyond Messages, Google is expected to extend protections system‑wide across Android and its hardware partners, including in‑notification scam detection, automatic screen‑sharing privacy by hiding sensitive data, call forwarding protection to prevent back‑channel hijacks, and a bank impersonation check that can automatically hang up if verification fails. The company cites Brazil as a live example where bank‑impersonation checks have blocked over 3 million fake calls, with ambitions to broaden to further regions.