Google is rolling out a new set of scam defenses for Android, and the biggest piece is an opt-in “enhanced protection” setting in Google Messages that sends suspicious chats to cloud-based Gemini models for analysis. According to The Verge AI, the feature will arrive in the coming weeks. It builds on the Scam Detection tool that already runs on the device. Google also announced protections for call forwarding and screen-sharing, and it’s bringing scam alerts for chat notifications to phones from many more manufacturers.
The reason for the cloud step is simple. Today’s most damaging text scams don’t start with an obvious phishing link. They start with “wrong number” small talk that builds trust over days or weeks before the scammer asks for money. On-device models catch the obvious stuff. Google is betting that a bigger cloud model can spot that slower manipulation.
What Android Is Shipping
- Cloud-powered scam analysis in Google Messages. With enhanced protection turned on, conversations with unknown senders can go to Gemini models in the cloud, which judge what the sender is really after. “Scammers often disguise their true motives behind slow-burn conversations,” Google said in its announcement. The company describes this as something that “augments and extends our existing on-device detection, allowing us to identify novel and sophisticated attacks.”
- Encrypted, walled-off processing. Google says user data stays encrypted during the analysis and is “processed in a secure dedicated space” that nobody but the user can access, Google included. That promise matters most here, because the setting sends private messages off the phone.
- Link checks with Gemini Nano. The same mode uses Gemini Nano, Google’s on-device model, to check real-time web signals on unfamiliar URLs before you open them. The aim is to flag a malicious link before the page ever loads.
- Call-forwarding lockdown. Android will stop apps from turning on call forwarding and will warn you when something tries. It’s a quiet fix for a nasty attack. Scammers who can redirect your calls can grab one-time passcodes and verification calls meant for you.
- Automatic hiding during screen-share. Sensitive information will now be hidden automatically when you share your screen in Chrome and Android WebView. That goes straight at the “tech support” scam, where a fraudster talks the victim into sharing their screen and then reads off passwords or banking details.
- Scam alerts beyond Pixel. Scam detection for chat notifications, which launched on Pixel phones last year, is expanding to Honor, Motorola, Oppo, OnePlus, Realme, Tecno, Vivo, Xiaomi, and others. Google says it plans to reach “hundreds of millions of additional users later this year.”
Why This Matters
What stands out is the hybrid design. Small on-device models handle fast, private checks like links. A bigger cloud model takes the harder job of reading intent across a whole conversation. That’s the same split showing up across the industry, as companies figure out where on-device AI is good enough and where it isn’t.
The expansion to other phone makers may matter more than any single feature. Pixel is a small slice of the Android market. Taking scam detection to Xiaomi, Vivo, Oppo, and Tecno puts it on phones in many of the regions where text fraud does the most harm. Protection that only works on flagship phones protects very few people.
Caveats Worth Noting
- It’s opt-in. Enhanced protection won’t be on by default, so the people most at risk may never switch it on.
- Privacy is a tradeoff. Encryption and an isolated processing space are reassuring, but some users won’t want any private messages leaving their phone.
- Timing is loose. The Messages feature is due “in the coming weeks,” and the wider rollout to partner phones is set for “later this year.”
- No word on pricing. The announcement doesn’t mention a cost, and these features look like built-in Android protections, not a paid add-on.
What Comes Next
Scammers adapt fast, and AI is now helping both sides. Google’s move tells you defense is shifting from blocking known bad links to reading intent in real time. Expect Apple and messaging platforms like WhatsApp to feel pressure to match it. The full rundown is available at The Verge AI.