Google announced an opt-in enhanced protection mode for Google Messages on 14 October that will use cloud Gemini models to assess possible scams in conversations with unknown senders. The setting will roll out in the coming weeks, according to Google’s announcement, and will add cloud analysis to the app’s existing on-device Scam Detection.
Key points
- Enhanced protection will use cloud Gemini models through Private AI Compute to assess conversations with unknown senders.
- On compatible devices, Gemini Nano will check unfamiliar links against real-time web signals before they are opened.
- Android is extending scam alerts, screen-sharing privacy and call-forwarding safeguards beyond Messages.
Google Messages adds cloud analysis for unknown senders
Google says its current Scam Detection tools in Phone by Google and Google Messages flag more than 5 million suspected scam calls and texts each month. That figure counts items flagged across calls and texts. The new Messages setting is intended for conversations in which the suspicious part may emerge gradually, rather than arriving as an obvious request in a single text.
With enhanced protection enabled, Google says Messages can send complex conversational context for analysis by its cloud Gemini models. Think of the difference between checking one envelope for a familiar warning sign and reading a sequence of letters to understand where the correspondence is heading. Google says the cloud models are meant to assess a possible scammer’s intent when the sender is unknown, supplementing detection that already runs on the phone.
Google says that analysis takes place through Private AI Compute: the data is encrypted and processed in a dedicated secure environment that the company cannot access. The company presents this as a way to use larger cloud models while keeping the conversation private. Enhanced protection remains a setting that must be switched on, rather than a replacement applied to every Messages conversation.
The distinction between processing on the phone and in the cloud matters to the way this feature is being built. GovTech and Google DeepMind’s work to develop on-device AI scam detection takes the on-device route. Google’s Messages update retains that kind of local detection while adding an optional route to cloud models for more involved conversations.
Google and YouGov surveyed 4,500 smartphone users across the United States, Canada and Australia. Google says 40% described themselves as very or extremely concerned about mobile fraud. The company’s separate monthly figure concerns suspected calls and texts flagged by its tools, rather than the number of people in that survey who encountered a scam.
Gemini Nano checks unfamiliar links
The enhanced setting also changes how Messages examines links. Google says its existing URL protection flags known malicious and phishing sites, while scammers can create addresses that have not entered those lists. On-device Gemini Nano will combine real-time web signals with a check of an unfamiliar link before it is opened. Redesigned warnings will also give more context before a link from a number outside the contacts list is tapped.
Opening a link in a text from an unfamiliar number could therefore involve a warning before the page loads, including when its address is too new for a list of known malicious sites. That would depend on the optional setting being enabled and, for the Gemini Nano check, on having a compatible device.
The Gemini Nano link check is available on the Pixel 10 and later, the Galaxy S26 series, Galaxy Z Flip 8 and Z Fold 8, and other select devices, 9to5Google reported. The revised warnings had previously been limited to devices in select countries or those with Android Advanced Protection enabled, Google says. For companies building Android phones, support for the new link check therefore depends on more than distributing the Messages app.
Android extends alerts beyond Google Messages
Google is also expanding its on-device scam alerts for chat notifications beyond Pixel phones. These alerts identify suspicious messages in the notification shade before a chat app is opened. Google says partnerships with Android device makers will bring the protection to hundreds of millions of additional users later in 2026.
9to5Google names HONOR, Infinix, iQOO, itel, JOVI, Motorola, OPPO, OnePlus, realme, TECNO, vivo and Xiaomi among the participating makers. The Verge lists Honor, Motorola, Oppo, OnePlus, Realme, Tecno, Vivo and Xiaomi.
During a live screen-share session, Android’s automatic privacy protection will blank out login fields and payment screens in Chrome and Android WebView, extending a safeguard Google says already operates in native apps. A separate call-forwarding measure addresses another route for fraud: Google says scammers can redirect incoming calls to obtain one-time passcodes, and Android blocks apps from enabling forwarding in the background. It also warns before a forwarding code entered manually on the dial pad takes effect.
Google will make the call-forwarding protection available from early December on Pixel and other select devices, 9to5Google reported.
