Google is doubling down on its fight against online scams by integrating its on-device large language model, Gemini Nano, across Chrome and Android. The move comes as part of a broader initiative outlined in Google’s newly released Fighting Scams in Search Report, where the tech giant detailed how artificial intelligence is now a core pillar in its scam prevention strategy.
Over the past decade, Google has invested heavily in AI to enhance security across its platforms. By analyzing large volumes of web data, the company can now detect and neutralize coordinated scam campaigns with greater precision. According to Google, this has led to a twentyfold increase in the number of scammy pages identified and blocked, ensuring safer and more reliable search results.
A major advancement is the rollout of Gemini Nano to Chrome, where it functions as an on-device safeguard against phishing and scam sites. Google claims this offers twice the level of protection compared to its standard mode, with the added benefit of being able to flag new scam types that haven’t previously been seen. This is particularly effective for threats like remote tech support scams, which are growing in sophistication.
On Android, Google is also enhancing user safety with AI-powered notifications in Chrome. If a machine learning model detects suspicious content, users receive a warning with the option to block or allow future notifications. This empowers users to manage potentially malicious messages in real time.
Further bolstering defenses, the company has launched AI-powered scam detection within Google Messages and the Phone by Google app. These tools are designed to intercept scams that originate from seemingly innocuous calls or texts but quickly escalate into high-risk scenarios.
In their blog post, Google emphasized that LLMs like Gemini Nano are crucial for adapting to the ever-changing landscape of scams, allowing for faster detection and more contextual risk analysis. By embedding AI directly into devices, Google not only improves protection but also ensures user privacy by processing data locally.