Google announced a sweeping set of AI-powered scam defenses for Android on October 14, led by a new opt-in Enhanced protection mode in Google Messages that enlists the company's cloud Gemini models to assess suspicious texts from unknown senders. The company says its Scam Detection tools already flag more than 5 million suspected scam calls and texts every month, and the new features are designed to catch more sophisticated attacks before users engage with them.
The announcement, laid out in an Android Security blog post by Google Messages senior product manager Alberto Pastor Nieto and Android product manager Shaifali Bhandari, bundles messaging defenses with expanded scam alerts in chat notifications, screen-sharing privacy for Chrome, call-forwarding safeguards, and a verified financial calls program aimed at bank impersonation scams. For more context on this story, see our ongoing latest AI developments.
Enhanced Protection Mode: Cloud Gemini for Unknown Senders
Enhanced protection is a new opt-in security setting in Google Messages that will roll out in the coming weeks. With the setting enabled, Scam Detection can draw on Google's cloud Gemini models through Private AI Compute to assess complex message context and identify a scammer's intent earlier in conversations with unknown senders. Google says the cloud analysis supplements its existing on-device detection and is built to surface novel and sophisticated attacks.
Privacy is the obvious sticking point for shipping message content to the cloud, and Google's answer is its Private AI Compute infrastructure: messages processed there are encrypted and handled inside a dedicated secure environment that remains inaccessible to anyone else, Google included, according to the company. A separate footnote notes that users can still optionally report a message as spam, which does send the message and its contents to Google for analysis.
The cloud-powered analysis requires capable hardware: it will be available on Pixel 6 and newer, the Galaxy S24 series and newer, Galaxy Z Flip6 and Z Fold6 and newer, and other Android devices with AICore support.
On-Device Gemini Nano Checks Links in Real Time
A second layer of defense runs entirely on the phone. On-device Gemini Nano weaves real-time web signals into link safety, checking whether an unfamiliar URL is trustworthy before the user taps it — a guard against attackers who routinely register brand-new websites that existing blocklists have not yet cataloged. These Gemini Nano link checks require the newest hardware: Pixel 10 and later, the Galaxy S26 series, Galaxy Z Flip8 and Z Fold8, and other select devices.
A third layer applies more broadly. Google is redesigning warnings that slow users down with added context before they open links arriving from numbers outside their contacts, a checkpoint previously limited to devices in select countries or those running Android Advanced Protection. The redesigned warnings will reach any device running Android 6 or later.
Expanding Scam Alerts Across the Android Ecosystem
Beyond Google Messages, the company is partnering with Android device makers to expand Scam Detection in chat notifications — a feature that launched on Pixel phones and flags suspicious messages directly in the notification shade before the user even opens the chat app. Google's footnotes name HONOR, Infinix, iQOO, itel, JOVI, Motorola, OPPO, OnePlus, realme, TECNO, vivo and Xiaomi as participating manufacturers, with availability varying by brand. Google says the on-device protection will reach hundreds of millions of additional users later in 2026.
Android is also extending its automatic hiding of sensitive information during screen sharing. The feature, which already obscures login fields and payment screens in native apps during an active screen-share session, is being extended to Chrome and Android WebView, and rolls out during October 2026 to devices running Android 15 or newer.
Why Google Is Moving Now
The scale of the problem justifies the breadth of the response. Phone and text scams have industrialized in recent years, with criminal operations using AI tools to draft convincing lures, clone voices and automate targeting at scale. Google's own numbers — 5 million suspected scams flagged per month — suggest the volume is far beyond what human review or static filters can handle, which is why the company is layering large models into the detection path at both the cloud and on-device tiers.
The opt-in design of Enhanced protection will draw scrutiny from both directions: privacy advocates will want to verify Google's claims about Private AI Compute isolation, while some security researchers argue protective defaults do more good than opt-in settings, since the users most likely to fall for scams are the least likely to change settings. Still, the tiered rollout — cloud analysis for capable devices, on-device Nano checks for the newest flagships, and basic link warnings for anything running Android 6 or later — gives Google a way to raise the baseline for nearly every active Android phone.
For users, the practical takeaway is simple: when the setting appears in Google Messages in the coming weeks, enabling Enhanced protection hands Gemini both the cloud and on-device models as screening tools for messages from strangers. Given that Google reports flagging millions of scams monthly with current tools, the incremental coverage could matter most in exactly the conversations that look clean to today's filters.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
