TechFlow News, May 6: A paper released by researchers from the University of Cambridge, the University of Edinburgh, and the University of Strathclyde shows that generative AI has not significantly transformed hackers into “super hackers.” The research team analyzed 97,895 posts on cybercrime forums following ChatGPT’s launch in November 2022 and manually reviewed over 3,200 of those posts. They found that 97.3% of the sample was categorized as “Other”—i.e., not actual discussions about using AI for criminal purposes—while only 1.9% involved tools such as “vibe coding.”
The study notes that so-called “Dark AI” tools like WormGPT and FraudGPT generate more market hype than practical malicious software; many forum posts instead focus on seeking free access, guessing how the tools work, or complaining about their ineffectiveness. The study concludes that AI’s currently observable applications in crime are largely confined to low-barrier, high-frequency activities—including mass SEO spam, romance scams, voice cloning, image generation, and low-cost AI-generated nude image services.
