Scammers will do anything to take your money from you, and it seems there’s a new scam each and every day. By now, you’re probably all too familiar with how scammers are vying for a piece of your pay packet, with email and SMS clearly up there, but also call-back scams are something else we’re seeing.
But phishing sites are one of the biggest problems, as scammers build fake versions of real places to trick you. The links can come over text or email, and you can even make your way there from social ad campaigns or links from friends who’ve fallen down the rabbit hole, and are now sending it to you.
Search, however, is almost always somewhat immune. The problem is fake sites are typically competing against the real website, and will lose out in Google, Bing, and nearly every other engine.
And yet AI may be a different story altogether.
While search and AI often have what they find in common, AI chatbots may more easily find websites that it believes are real solely on content, and that can be a problem. In the age of ghost stores, it could end up being a major issue for anyone, young or old.
Norton’s Gen Threat Labs has chimed in with that recently, noting a 54 percent increase in malicious content served to chatbots, with both ChatGPT and Claude at risk, and potentially opening a new reason to consider internet security.
It might sound like something that should be further in the horizon, especially with AI chatbot use for search still somewhat young, Norton has confirmed that fake websites are being found in chat. Frustratingly, there’s next to no way to work out whether a site is fake from its AI chatbot presence, and you might just click on it being none the wiser.
“Unfortunately, this is already happening,” said Jakub Kroustek, Director of Threat Research and Applied AI at Gen Threat Labs.
“The issue is not AI acting maliciously on purpose; it’s that chatbots aren’t built to spot scams,” he said. “Nobody has to trick the chatbot, it recommends the site on its own.”
The chatbot essentially is seeing fake pages as part of its search and indexing, and could be learning them from training data. In one supplied example, Norton noted the presence of a fake shop, essentially acting as a ghost store designed to lure victims.
While Pickr couldn’t verify with the same query, it did find at least one hallucination in the same localised version of the query, while repeated tests uncovered at least one shop that appeared on the surface as a ghost shop. Similar searches on Google didn’t quite uncover the same pages, though searching without location did find at least one initially.
It seems that scammers are beginning to make a dent, and while search is normally on top, AI chat systems may be serving these up, a situation Norton says may include phishing sites and other forms of internet scams, including investment scams and fake download sites.
“These sites get into AI answers in three ways: the chatbot finds them through web search while answering, it learned them from training data, or it invents a plausible web address that criminals later register,” said Kroustek to Pickr.
If you are using AI chat services as a search engine, internet security could end up becoming a potential solution to at least go on the defensive, providing some protection against dodgy websites where AI has missed the memo.
It clearly isn’t the only way to go on the defensive, and being suspicious of any link (much like you might with a phishing email) could be one way to prevent being ensnared as a victim.
However, it does signal that trust in what you see from search might just be that little bit more complicated, especially if you lean on AI to do any of your searching for you.
