Google brings Gemini Nano scam protection to Chrome and Android

Ever searched for an airline customer service number only to realize you dialed a scammer? Google knows it is happening, and it is deploying generative AI to fight back.

Google dropped a detailed report outlining how it is overhauling scam defenses across Search, Chrome, and Android. The biggest shift is bringing small, on-device AI directly to your browser to catch fraud in real time.

Search scam detection jumps 20x

Google says improvements to its AI detection systems and classifiers have enabled Search to catch 20 times the number of scam pages compared to before, blocking hundreds of millions of shady search results every single day.

One specific target has been fake airline customer support scams, where travelers looking for urgent flight help get tricked into dialing fraudulent call centers. Google claims its updated AI systems have already reduced these specific scams in Search by more than 80%.

Instead of just matching keywords, the models analyze vast swaths of web text to spot coordinated campaigns and sniff out zero-day scam setups before they gain traction.

Gemini Nano moves into desktop Chrome

Perhaps the most practical upgrade is happening inside desktop Chrome. Users with Enhanced Protection enabled are getting an extra layer of security powered directly by Gemini Nano, Google's lightweight on-device LLM.

Running locally means Chrome can instantly parse complex websites to understand their underlying intent without sending raw browsing data back to a server. Gemini Nano is currently tasked with stopping remote tech support scams—the kind that flash loud fake virus warnings urging you to grant remote access.

Meanwhile, Chrome on Android is getting its own on-device machine learning tool to suppress spammy notification barrages, offering an instant unsubscribe button. Coupled with the existing AI scam detection in Google Messages and the Phone app, Google is trying to box scammers out of the entire OS.

Why it matters

For everyday users, this is passive protection that actually works. You do not need to learn the latest social engineering tactics if your browser and phone can flag them on the fly. Turning on Enhanced Protection in Chrome is an easy win, especially on devices used by less tech-savvy family members who are frequently targeted by fake tech support popups.

Across the industry, this signals a major shift from reactive blacklist databases to proactive semantic understanding. Scammers historically churned through thousands of disposable domains to bypass URL filters. An on-device model like Gemini Nano disrupts that playbook because it reads the page just like a human would, identifying deceptive behavior regardless of the URL.

We think using lightweight models for on-device security is one of the smartest applications of small LLMs yet. They do not need to write poetry; they just need to recognize shady patterns quickly. The slight catch is user trust. Even with on-device processing, having an LLM actively scan every page you browse under Enhanced Protection will inevitably raise eyebrows among privacy purists. Google will need to be transparent about boundaries as it expands Gemini Nano across Android.