Almost anyone who has used the internet has probably experienced that alarming moment when a window pops up claiming your device has a virus, encouraging you to click for tech support or download security software. It’s a common online scam, and one that Google is aiming to fight more aggressively using artificial intelligence.
Google says it’s now using a version of its Gemini AI model that runs on users’ devices to detect and warn users of these so-called “tech support” scams.
It’s just one of a number of ways Google is using advancements in AI to better protect users from scams across Chrome, Search and its Android operating system, the company said in a blog post Thursday.
The announcement comes as AI has enabled bad actors to more easily create large quantities of convincing, fake content — effectively lowering the barrier to carrying out scams that can be used to steal victims’ money or personal information. Consumers worldwide lost more than $1 trillion to scams last year, according to the lobbying group Global Anti-Scam Alliance. So, Google and other organizations are increasingly using AI to fight scammers, too.
Phiroze Parakh, senior director of engineering for Google Search, said that fighting scammers “has always been an evolution game,” where bad actors learn and evolve as tech companies put new protections in place.
Related article ‘My job is to worry’: Google DeepMind COO Lila Ibrahim on the responsibility of building AI
“Now, both sides have new tools,” Parakh said in an interview with CNN. “So, there’s this question of, how do you get to use this tool more effectively? Who is being a little more proactive about it?”
Although Google has long used machine learning to protect its services, newer AI advancements have led to improved language understanding and pattern recognition, enabling the tech to identify scams faster and more effectively.
Google said that on Chrome’s “enhanced protection” safe browsing mode on desktop, its on-device AI model can now effectively scan a webpage in real-time when a user clicks on it to look for potential threats. That matters because, sometimes, bad actors make their pages appear differently to Google’s existing crawler tools for identifying scams than they do to users, a tactic called “cloaking” that the company warned last year was on the rise.
And because the model, called Gemini Nano, runs on your device, the service works faster and protects users’ privacy, said Jasika Bawa, group product manager for Google Chrome.
As with Chrome’s existing safe browsing mode, if a user attempts to access a potentially unsafe site, they’ll see a warning before being given the option to continue to the page.
In another update, Google will warn Android users if they’re receiving alerts from fishy sites in Chrome and let them automatically unsubscribe, so long as they have Chrome website notifications enabled.
Google has also used AI to detect scammy results and prevent them from showing up in Search, regardless what kind of device users are on. Since Google Search first launched AI-powered versions of its anti-scam systems three years ago, it now blocks 20 times the number of problematic pages.
“We’ve seen this incredible advantage with our ability to understand language and nuance and relationships between entities that really made a change in how we detect these scammy actors,” he said, adding that in 2024 alone, the company removed hundreds of millions of scam search results daily because of the AI advancements.
Parakh said, for example, that AI has made it better able to identify and remove a scam where bad actors create fake “customer service” pages or phone numbers for airlines. Google says it has has now decreased scam attacks in airline-related searches by 80%.
Google isn’t the only company using AI to fight bad actors. British mobile phone company O2 said last year it was fighting phone scammers with “Daisy,” a conversational AI chatbot meant to keep fraudsters on the phone, giving them less time to talk with would-be human victims. Microsoft has also piloted a tool that uses AI to analyze phone conversations to determine whether a call may be fraudulent and alert the user accordingly. And the US Treasury Department said last year that AI had helped it identify and recover $1 billion worth of check fraud in fiscal 2024 alone.