Live monitoring Updated 72-hour alert levelMajor alert Editor: Frédéric LENQUETTE
CYBERIACKCybersecurity & AI intelligence · by Frédéric LENQUETTE
AI & securityGood to know

Scam texts: Android offers to let Gemini read your chats with strangers

On October 14, Google announced new Android scam protections: AI analysis of conversations and blocking of booby-trapped call forwarding. Here is what is optional, and what Google leaves unsaid.

Translated from French by AI. The original French article is the reference version.

AI-generated illustration

A text from an unknown number, a link to a site created the day before, then a conversation that drags on to build trust. On October 14, Google announced a new protection mode for its Messages app, which hands these exchanges to its Gemini AI to spot a scam.

This “enhanced protection mode” is off by default: users must turn it on. Google says it will arrive “in the coming weeks.” It is the centerpiece of a set of new defenses presented on the Android security blog.

Do your messages go to the cloud?

Partly, yes. Conversations with unknown senders are analyzed by Gemini models hosted in the cloud, in an environment called Private AI Compute. Google says the data is encrypted there and inaccessible to anyone, Google included. The announcement does not mention any public audit of this promise.

Another part of the work happens on the phone: Gemini Nano, the small on-device model, assesses unknown links using real-time web signals before they are opened, on Pixel 10 and later, Galaxy S26 and a few other models. Messages will also show warnings before a link sent by a number not in your contacts is opened, starting with Android 6.

The other protections announced

  • Notification alerts: already available on Pixel, scam detection in messaging app notifications is due to reach HONOR, Motorola, OnePlus, OPPO, Xiaomi and other manufacturers “later this year.”
  • Screen sharing: from Android 15, login and payment fields are hidden during screen sharing in Chrome and web views, a defense against fake advisers who ask you to share your screen.
  • Call forwarding: in early December, on Pixel and a few other devices, Android will stop apps from quietly enabling call forwarding and will warn users before a manually typed forwarding code takes effect.
  • Verified bank calls: Android checks that a call claiming to come from the bank really does, and hangs up otherwise. Google claims more than 3 million spoofed calls blocked in Brazil since June. Revolut is among the compatible institutions.

What about France?

That is the question the announcement leaves open. Google publishes no list of countries or languages for most of these features. Only Brazil is named, for bank calls. Before you talk to your teams or your family about it, check what actually appears in the Messages settings.

Phishing operations, such as the one Microsoft has just disrupted, send huge volumes. AI helps with sorting, but the basic rule does not change: a bank never asks you for a code over the phone, and a link from a stranger is not to be opened.

Method: AI-assisted monitoring, verified sources, article reviewed and approved by Frédéric LENQUETTE.

Share this article