Google announced on 14 October a package of scam defences for Android. The main new feature is Google Messages’ advanced protection mode: an optional setting that sends conversations with unknown senders to Gemini models in the cloud, to work out whether someone is trying to con you. It will arrive “in the coming weeks”.
How the advanced mode works
Until now, scam detection in Messages has worked mainly on the phone. With advanced protection, which is on only if you choose it, the analysis can draw on the most capable Gemini models running in the cloud. The idea is to catch scams that are built up exchange by exchange, where the fraudster doesn’t write the suspicious line straight away, and not just messages containing known keywords.
Google says the processing takes place in Private AI Compute, a protected space that, according to the company, nobody can access, not even Google. That is a guarantee given by Google itself: anyone who cares a great deal about the privacy of their messages would do well to weigh it up before switching the option on.
The mode also includes two link checks:
- Verification with Gemini Nano, directly on the device: it assesses little-known links, including brand-new sites that are not yet in security databases.
- Warnings before you tap: a redesigned warning appears on links sent from numbers that aren’t in your contacts.
The other new features
The rest of the announcement covers the whole system:
- Scam warnings in notifications for chats: they appear in the notification shade, first on Pixel. By the end of the year they will reach other brands, including Motorola, Xiaomi, OnePlus, OPPO, realme and vivo, with timing depending on the manufacturer.
- Privacy during screen sharing: in Chrome and Android WebView, on Android 15 and later, login and payment fields are hidden automatically.
- Call-forwarding block: it stops apps from switching it on in the background and warns you if you are being tricked into dialling a forwarding code. It starts in early December on Pixel and some other devices.
- Verified bank calls: with partner banks’ apps, Android checks in real time that the call really comes from the institution and, if it doesn’t, hangs up and warns you. Google says it has blocked more than 3 million spoofed calls claiming to be from Banco do Brasil and Itaú Unibanco since June. The service is live in Brazil and with Revolut.
Which phones are covered
Not every feature is for everyone:
- Cloud analysis: Pixel 6 and later, Galaxy S24 and later, Galaxy Z Flip and Fold 6 and later, plus other phones with AICore.
- Link verification with Gemini Nano: Pixel 10 and later, Galaxy S26, Z Flip 8 and Z Fold 8, and a few other models.
- Warnings on links from unknown numbers: Android 6 and later.
Google warns that some features may not be available in every country and has not published a feature-by-feature list. It is therefore unclear, for now, when and whether each one will reach Italy or Switzerland. Italian banks are not among the partners named for verified calls.
What we think
The right front to fight on is conversational scams: the fake “hi mum, I’ve changed my number” or the bogus courier can’t be stopped with a list of banned words. A model that reads the whole conversation has a better chance of spotting them, and it makes sense that the option is left to the user’s choice, given that it means sending messages off the phone.
Two reservations remain. The figures on effectiveness come from Google and from the assessment by Leviathan Security Group that it commissioned, which puts Android’s protections ahead of iOS’s: an interesting data point, but not an independent test. And coverage is patchy: anyone with a phone that is a few years old, or from a brand not on the list, will have fewer tools.
In the meantime, the usual rules apply: don’t click links from unknown senders, and never share your screen with someone who calls you “from the bank”. If you use Messages, it’s worth keeping an eye on the service’s settings over the coming weeks to find the new switch.
Written by the SpazioiTech newsroom with the help of AI tools, from facts verified across at least two independent outlets, and checked before publication. Spotted a mistake? Tell us.
Translated from the Italian original. Read in Italian
