A new report suggests that a significant portion of web traffic is courtesy of AI, and in the future, a majority of online activity could be artificial in nature. But this only hints at the larger (and potentially catastrophic) problem – using AI to scrape and harness the dark web.
According to TollBit, a company dealing in web-scraping activity, “The majority of the internet is going to be bot traffic in the future. It’s not just a copyright problem, there is a new visitor emerging on the internet.”
Tollbit adds that 1 of every 31 visits to its customers’ websites was from an AI scraping bot.
That includes AI chatbots, generative AI, and other more sophisticated programs data crawling for training purposes but also to retrieve real-time info (like product prices, tech specs, etc.).
Traditionally, websites try to limit AI bots’ access to sensitive and potentially copyrighted data, but the bots are getting more sophisticated. And this relatively benign AI subterfuge is child’s play compared to its usage on the Dark Web.
For those not in the know, the Dark Web comprises a small part of the Deep Web, unindexed content that makes up 90-96% of the entire internet. Accessible only through special web browsers, the Dark Web houses the most illegal, explicit types of content, along with a decentralized, virtual marketplace for illicit products (and services).
A good portion of Dark Web wares is highly-privileged proprietary data, and that’s where AI comes in.
A number of potential targets – law enforcement, governments, and companies – are using AI to monitor the dark web, searching for stolen classified data or potential crimes. Why waste time on tedious sting operations when AI can do the job faster and far more efficiently?
But it gets far more malignant than that. We’ve already seen chatbots repurposed for illegal activity, but the dark web also hosts highly-illicit offshoots of ChatGPT like WormGPT and FraudGPT. WormGPT was officially shut down recently, but interested parties can apparently still purchase access.
FraudGPT offers functionality like writing malicious code, creating phishing pages and hacking tools, and lots more. DarkBERT has similar features, but its training has taken place almost entirely on the dark web.
And there’s an even scarier prospect – using AI to cut through the phony criminal solicitations on the dark web and arrive at the genuine article. No definitive data exists about the percentage of legitimate dark web criminal ads, but AI could sift through the ads at superhuman speed and connect illicit buyers and sellers.
We’re all concerned about the existential threat of killer robots, but the far more immediate danger could be dark web AI.