There were about 3,000 posts on dark web forums and cybercrime Telegram channels in 2023 discussing artificial intelligence (AI) tools and how to exploit large language models (LLMs), like ChatGPT, for illegal purposes, according to a new report from Kaspersky.
Discussions on how to use AI for cybercrime peaked in March 2023 — coinciding with the release of OpenAI’s ChatGPT-4.
There were also 3,000 posts about ChatGPT accounts. While some of these were advertising stolen premium ChatGPT accounts for sale, others were services offering to create ChatGPT accounts in bulk using fake details. One forum member even offered free ChatGPT accounts, “presumably” stolen using malware, Kaspersky said.
“Threat actors are actively exploring various schemes to implement ChatGPT and AI,” Alisa Kulishenko, digital footprint analyst at Kaspersky, said. “Topics frequently include the development of malware and other types of illicit use of language models, such as processing of stolen data, parsing files from infected devices, and beyond.”
How Cybercriminals are Using AI Technology
Kaspersky’s report highlighted different ways threat actors use AI technology. One post discussed using OpenAI’s ChatGPT to create polymorphic malware. This new breed of malware is difficult to detect due to its ability to change its code while maintaining its core functions.
Though researchers “have not yet detected any malware operating in this manner,” it could appear in the future, the report said.
Kaspersky researchers also found cybercriminals using AI to analyze and exploit stolen data, significantly lowering the barrier to cybercrime entry. “Actions that previously required a team of people with some experience can now be performed at a basic level even by rookies,” the researchers said.
Interestingly, AI has even been integrated into cybercriminal dark web forums to produce automated advice on illegal activities. The report highlighted malicious ChatGPT-like models, like WormGPT, XXXGPT, and FraudGPT.
“WormGPT is currently the most widespread among those projects,” while others like WolfGPT and DarkBERT are also popular, Kaspersky said. In addition, researchers observed the misuse of legitimate open-source and cybersecurity penetration testing tools for cyber attacks.
“While AI tools themselves are not inherently dangerous, cybercriminals are trying to come up with efficient ways of using language models, thereby fueling a trend of lowering the entry barrier into cybercrime and, in some cases, potentially increasing the number of cyberattacks,” Kulishenko added.
AI Probably Won’t Revolutionize the Attack Landscape in 2024: Kulishenko
Kulishenko said AI and chatbots will probably not revolutionize the cyber attack landscape “at least in 2024.” Nevertheless, staying informed about this threat is key, especially for the corporate world, Kulishenko added. Kaspersky recommends that businesses adopt strong cybersecurity strategies to combat AI-driven cybercrime.
For individual netizens, we recommend bolstering your digital security by using one of our top-rated antivirus suites to protect your system from malware and other online threats. It’s also important to regularly update your software and operating systems and practice proper cyber hygiene.
You can protect your ChatGPT account and other accounts with a strong password. Refer to our password guide to learn how to create bulletproof credentials for your online accounts.
For more news, follow us on X (Twitter), Threads, and Mastodon!
