Google identifies online scammers with AI
Google Announces On-Device AI Model to Detect and Warn Users About Online scams
Google has announced the deployment of a version of its artificial intelligence (AI) model that runs directly on users’ devices to identify online fraud and alert them.
According to a report by the International Economics section of Webangah News Agency, citing Tasnim News Agency and CNN, nearly everyone who uses the internet has likely encountered the alarming moment when a pop-up suddenly appears, claiming their device is infected with a virus and urging them to click for technical support or download security software. This is one of the most common online scams, and Google now aims to combat it more aggressively using AI.
In a blog post on Thursday, Google revealed it has implemented a version of its AI model called Gemini, which operates directly on users’ devices to detect these so-called “tech support” scams and issue warnings.
This is just one of several methods Google employs across Chrome, Google Search, and the Android operating system to enhance user protection against fraud.Google’s announcement comes as AI now enables malicious actors to generate vast amounts of convincing fake content, increasing online scam risks while lowering barriers for execution. According to the Global Anti-Scam Alliance, users worldwide lost over $1 trillion last year due to such scams.
Google’s Senior Engineering Director for Search stated: ”Fighting scammers has always been an ongoing evolutionary race—where criminals adapt their tactics in response to tech companies’ efforts to implement new safeguards.” Speaking with CNN, they added: “Now both sides have new tools. The question is: Who deploys them more effectively? Who acts faster?”
While google has long used machine learning (ML) for service protection, recent advancements in AI have improved language understanding and pattern recognition—enabling faster, more accurate scam detection. In Chrome’s desktop “Enhanced Protection” mode, an on-device AI model can scan web pages in real-time upon loading for potential threats—a critical feature since scammers often design pages differently for search crawlers than what users see (a tactic called cloaking, which Google warned about last year).Chrome Group Product Manager Jessica Bawa explained that becuase Gemini Nano runs locally on devices:
- Services are delivered faster
- User privacy remains intact
