Google Messages adds opt-in cloud AI to screen unknown Android senders
The new Enhanced Protection mode uses Gemini models to detect scams and evaluate links before users tap them, available in the current app beta.
By Petra Murphy ·
Google is rolling out an Enhanced Protection mode for its Android messaging app to identify spam and scams. This new tier supplements the existing Standard Protection, which currently blocks known spam and harmful links.
The primary new feature is Scam Detection. It employs cloud-based Gemini models to analyze the context of message threads from unknown senders. Google states this method identifies the intent of scammers more quickly than on-device detection, particularly for novel attacks. To maintain privacy, the system processes messages within Android's Private AI Compute environment. In this secure space, data is encrypted and inaccessible to Google or other parties.
A second component is enhanced link protection. This tool combines the on-device Gemini Nano model with real-time web signals to verify unfamiliar links before a user taps them. The feature is limited to specific hardware, including the Pixel 10 and newer, the Samsung Galaxy S26 series, the Galaxy Z Flip 8, and the Galaxy Z Fold 8.
The update also changes how warnings appear. Messages from unsaved numbers now trigger alerts that provide users with more context and time to pause before interacting with a link.
Enhanced Protection is opt-in. Users access the setting via the Protection and safety menu in Google Messages. The interface now offers three distinct options: Enhanced protection, Standard protection, and No protection. The app labels the final option as not recommended. The feature has appeared in the latest beta version, identified as 20261020_00_RC00.
