Trending Articles
Trending Topics
Trending
Google has lifted the lid on how it’s using artificial intelligence to battle scammers.
Earlier this month, Google published a report detailing how the company has been using AI to prevent its searches returning results for scam pages as part of the ongoing arms race between the tech giant and online fraudsters.
According to Google, its AI-powered scam detection efforts now enable the company to filter out “hundreds of millions” of potentially fraudulent pages before they hit users’ search results every day.
Google first started using AI to tackle scam results three years ago, according to the report, and stepped up its AI efforts last year, using large language models (LLMs) and other AI tools to detect potential scam activity. LLMs have been crucial in picking up scams in languages other than English, Google said, and watching for when fraudulent activity in one language appears in another.
Google is using new AI models to detect “subtle linguistic patterns and thematic connections” that indicate when pages are harbouring scam activity, and flag up when online criminals refine their tactics or develop new scams.
Google is also putting its AI tools to work spotting subtle indicators of fraudulent sites. This includes detecting links between websites that suggest coordinated scam activity, even if the individual websites look legitimate.
It is spotting trends that online criminals are using to lure in would-be victims, too. For example, creating “dedicated protections” to combat recent rises in scam activity targeting areas such as airline customer services queries and government and visa services. This has decreased the amount of scams in both areas by 80% and 70% respectively.
The company’s AI-powered anti-scam efforts aren’t just being deployed across its search products, they’re also making their way onto devices too.
Google is now eating its own dog food when it comes to scam detection: it’s using Gemini Nano, its on-device LLM, to power Chrome’s Enhanced Protection mode, where it helps to flag up if a user is trying to visit a website that may actually be a scam site.
“If you turn on Enhanced Protection, much of the additional protection you receive comes from advanced AI and machine learning models designed to spot dangerous URLs engaging in known phishing, social engineering and scam techniques. These models are trained to distinguish between real and malicious websites based on millions of real-world examples,” Google said.
Gemini Nano is also being used in the fight against remote tech support scams, where criminals trick a user into thinking they have a problem with their PC to either extort money or gain remote access to the victim’s machine.
According to Google, tech support scam sites are often only online for short periods of time – as little as ten minutes – so might not have been crawled before. Gemini Nano will pick up signs of scam activity when a user navigates to a potential tech support scam site, looking for indicators the site is unsafe. it will then summarise the security signals and feed them into Safe Browsing, which will show a warning interstitial informing the user the site is likely fraudulent.
In future, Google plans to use Gemini Nano to detect other popular types of scams, such as unpaid toll and package tracking scams, as well as extend its capabilities to Android devices.
Chrome on Android has seen new AI-driven anti-scam functionality rolled out this month: according to Google, Chrome on Android now features an on-device machine learning model which can analyse the content of notifications sent to Android phones and flag up which notifications may be spam or fraudulent.
