Google Messages will add an opt-in “enhanced protection” mode that uses cloud-based Gemini models to evaluate conversations with unrecognized senders for signs of a scam. The feature is intended to catch slow-building fraud that may appear harmless in individual messages before asking for money or credentials later.
Google says conversation data sent for analysis is encrypted and processed in a dedicated environment inaccessible to both the company and other parties. On the phone, Gemini Nano will separately inspect current web signals for unfamiliar links before a user opens them. The cloud component extends existing on-device scam detection rather than replacing it.
The setting is due to roll out in the coming weeks. Google did not provide accuracy or false-positive rates, so users will need to judge whether the added cloud analysis is worth enabling. The privacy claim also depends on the implementation of the isolated processing environment, not encryption alone.
Android is adding related defenses outside Messages. Apps will be blocked from silently enabling call forwarding, and users will receive a warning when one tries. Chrome and Android WebView will hide sensitive information during screen sharing. Google is also expanding scam detection for chat notifications beyond Pixel phones to devices from Honor, Motorola, Oppo, OnePlus, Realme, Tecno, Vivo, Xiaomi and others, with hundreds of millions of additional users expected later this year.