Gemini Will Screen Your Texts: Google Rolls Out AI Scam Warnings for Android

By Marcus Bennett

Google will use its most advanced Gemini AI models to warn Android users about scam text messages, through an opt-in Scam Detection setting in the Messages app that the company says will roll out in the coming weeks. The feature, announced on November 6, 2025, evaluates a sender's intent and flags messages that may come from a bad actor, processing them inside Private AI Compute — an encrypted, secure space that Google says even it cannot access.

The announcement matters because text-based fraud has become one of the most common attack surfaces on mobile platforms, and Google is now placing on-device AI models, rather than blocklists alone, at the center of its defense strategy. The move also extends Gemini's role beyond assistant features into core operating system security.

How does the Gemini scam detection actually work?

Toggling on the new setting enables Scam Detection, which Google said taps its most advanced Gemini models to evaluate the sender's intent and warn the recipient if a message could be deceptive. In Google's own example, a sender pretended to show the recipient a photo of them with their dog. When told they had the wrong number, the sender kept engaging and asked for the recipient's name — a classic pattern of engagement scams.

The architecture has two layers:

• Private AI Compute: Android analyzes messages in an encrypted secure space inaccessible to anybody else, including Google itself.
• On-device link analysis: Gemini Nano, Google's smallest model designed to run locally on phones, will scan links in Messages to verify they are safe before the user taps them.

The on-device link checking will be limited, at least initially, to select hardware: Google Pixel 10 and later, the Samsung Galaxy S26 series, Galaxy Z Fold 8 and Galaxy Z Flip, among other devices Google did not name. Google has also redesigned its existing scam warnings to give users more context and extra time before tapping links sent from numbers not in their contacts.

Which devices get Scam Detection, and when?

Scam Detection already exists in chat notifications on Pixel devices. Google is now expanding availability to more manufacturers, and the notification-level version carries a practical advantage: it warns about potential scams inside the message notification itself, so the user does not even have to open the message.

The rollout partners include:

• Honor
• Motorola
• OPPO
• OnePlus
• realme
• vivo
• Xiaomi

Those manufacturers are expected to enable the feature later this year, Google said. No specific dates were given for individual vendors.

What is arriving beyond Messages?

Google bundled several other security additions into the same announcement. All devices running Android 15 and newer will receive automatic screen-sharing privacy later this month. When a user shares a screen containing Chrome or an Android WebView, the system will hide passwords, banking information and credit card security numbers by default.

In early December, Google will roll out call forwarding protection on Pixel and other select devices. The feature blocks apps from secretly enabling call forwarding in the background. Android will also warn users if a scammer tries to trick them into typing an MMI code — a short string of figures — that would switch on call forwarding on the device.

What about the three million blocked calls?

The most concrete performance figure in Google's announcement comes from Brazil. Google teamed up with banks there to combat bank impersonation, and Android now performs a verification check on incoming calls claiming to be from a bank. If the caller fails the check, Android hangs up the call and alerts the user.

Google reported that Android has blocked over three million fake calls since launching the feature in June. The company did not say if or when the bank verification tool will reach the United States, but it stated it will work with more partner banks in the future.

Why this signals a shift

The numbers frame the stakes: three million blocked calls in roughly five months from one feature in one market, against a backdrop of engagement scams that begin with a single unsolicited text. By pairing encrypted cloud analysis in Private AI Compute with on-device Gemini Nano link checks, Google is betting that AI intent evaluation — not static filters — becomes the default fraud defense on Android. Whether the Brazilian bank-verification model travels to other markets will be the test of how far Google can push carrier- and bank-dependent protections.