Google adds opt-in AI scam checks and sensitive-screen hiding to Android

Google is adding Android defenses designed to catch drawn-out text scams, hide sensitive information during screen sharing and stop apps from quietly redirecting calls. But the October 14 announcement is not a single upgrade available everywhere: its protections arrive on different schedules, and the new cloud-powered message checks require users to opt in.

Messages looks beyond the suspicious link

The central AI addition is Enhanced protection mode in Google Messages. Google says it lets Scam Detection use cloud Gemini models to examine complex conversations with unknown senders. The target is a slow-burn scam: an exchange whose purpose becomes clearer across several messages, rather than an obviously fraudulent opening text.

That cloud analysis supplements existing on-device detection. Google says it runs inside Private AI Compute, a dedicated processing environment where data is encrypted and inaccessible to others, including Google. This is the company’s privacy assurance for the optional analysis—not a claim that all message checking stays on the phone.

• With Enhanced protection enabled, on-device Gemini Nano uses real-time web signals to assess unfamiliar links before a user taps them, rather than relying only on known threats.
• Redesigned warnings add a checkpoint for links from numbers outside a user’s contacts. This safeguard was previously limited to select countries or devices with Android Advanced Protection enabled.

Google is also taking chat-notification scam alerts beyond Pixel. These on-device warnings appear in the notification shade before someone opens a supported chat app. Google says manufacturer partnerships will bring the protection to hundreds of millions of additional users.

Blocking the tools a scammer might misuse

The screen-sharing change extends an existing native-app protection to Chrome and Android WebView. Login fields and payment screens are automatically blanked out in the shared view, protecting sensitive moments such as entering a password or credit card security number.

Call-forwarding protection addresses another route into an account: redirecting incoming calls to intercept one-time passcodes. The safeguards block apps from enabling forwarding silently. They also warn before a manually entered forwarding code takes effect, covering situations where someone is persuaded to type it themselves.

Bank verification offers an earlier test

Google also highlighted verified financial calls, launched in June. Android checks with participating banks’ official apps whether the bank initiated a call. If a caller claiming to represent the bank fails verification, Android ends the call and alerts the user. Participants include Revolut, with more partners planned.

Google says it has blocked more than 3 million spoofed calls involving Banco do Brasil and Itaú Unibanco since launch. That total reflects calls impersonating inbound-only bank numbers—lines never used to contact customers—not a measured success rate for the new AI message checks.