In a concerning turn of events, cybercriminals have reportedly developed their own versions of large language models (LLMs), similar to OpenAI’s ChatGPT, to potentially amplify their criminal activities, such as crafting phishing emails or malware. These illicit chatbots have surfaced on dark-web forums and marketplaces since early July, raising questions about their authenticity and purpose.
These rogue LLMs, known as WormGPT and FraudGPT, claim to lack the safety measures present in legitimate LLMs from companies like Google, Microsoft, and OpenAI. The criminal developers assert that their systems can sidestep ethical barriers and generate text without limitations, enabling various illicit activities.
WormGPT, which allegedly builds on the GPTJ language model, is advertised as useful for phishing schemes, particularly targeting novice cybercriminals. It can create emails that are not only persuasive but strategically cunning, as demonstrated by a simulated business email compromise scam.
FraudGPT’s creator boasts that the chatbot can generate undetectable malware, locate leaks and vulnerabilities, and contribute to online scams. The system was offered for a subscription fee, suggesting that cybercriminals could access its capabilities for malicious purposes.
The authenticity of these criminal chatbots remains uncertain, and the cybercriminal ecosystem is known for scams and deception within its ranks. Experts acknowledge that cybercriminals often defraud each other, casting doubt on the legitimacy of these tools.
While some signs suggest the potential use of WormGPT, doubts persist about FraudGPT’s credibility. The broader concern underscores the ongoing exploitation of generative AI’s hype by scammers, prompting calls for heightened transparency, accountability, and oversight in the deployment of AI technologies.
As the situation unfolds, experts remain vigilant, assessing the risks posed by these illicit chatbots and advocating for safeguards to prevent their misuse for criminal activities.
