Android adds AI scam detection to Google Messages

Google is expanding AI-based scam defenses in Google Messages and adding new safeguards against mobile fraud on Android.

Google is expanding Android's defenses against phone and text scams, adding an AI-powered "enhanced protection" mode to Google Messages and several new safeguards across the operating system.

Enhanced protection in Google Messages

Google Messages already has a Scam Detection feature that runs on the device. The new enhanced protection setting, rolling out over the coming weeks, adds cloud-based Gemini models to analyze conversations with unrecognized senders. Google says scammers often hide their goals inside slow, friendly conversations, and that cloud analysis helps identify newer and more sophisticated schemes that on-device checks might miss.

According to Google, message data is encrypted during this analysis and processed in a secure, dedicated environment that Google itself cannot access. The mode is opt-in. It also uses the on-device Gemini Nano model to check real-time signals for unfamiliar links before you open them, to judge whether they are malicious.

Other new protections

Android will also block apps from turning on call forwarding and notify users when something tries to, a tactic scammers use to intercept verification calls. In addition, sensitive information will be hidden automatically while screen-sharing in Chrome and Android WebView, which helps against scams where a fake support agent asks you to share your screen.

Why it matters for builders

This is a good example of a hybrid AI design: a small on-device model handles fast, private checks, while a larger cloud model is used only when needed and only with explicit consent. Many products can use the same pattern to balance privacy, latency and accuracy.

It also changes the environment for legitimate apps. If your product sends text messages or links to users, expect stronger filtering. Use recognizable sender names, avoid URL shorteners that look suspicious, and keep message templates clear and consistent so automated systems do not mistake them for scams.

For apps that rely on call forwarding or screen sharing for legitimate features, check how the new restrictions affect your flows before they reach your users.

Gemini and Google's developer tools are listed in our directory if you want to experiment with similar on-device and cloud models.

Source: The Verge. This article is PickAI’s own summary and analysis based on the original reporting.