WormGPT: ChatGPT rival with ‘no ethical boundaries’ sold on dark web

Researchers have raised concerns about a powerful AI tool called WormGPT, which is being marketed on the dark web and poses significant risks for cyberattacks. Unlike leading AI tools like ChatGPT and Bard, which have safeguards against misuse, WormGPT has “no ethical boundaries or limitations” and is specifically designed for malicious activities.

According to cybersecurity firm SlashNext, WormGPT is a sophisticated AI model trained on diverse data sources, with a focus on malware-related information. It can generate human-like text that can be utilized in hacking campaigns. The researchers conducted an experiment with WormGPT, instructing it to create an email aimed at pressuring an unsuspecting account manager into paying a fraudulent invoice. The generated email was not only persuasive but also strategically cunning, showcasing the tool’s potential for sophisticated phishing attacks.

Screenshots provided by WormGPT’s anonymous developer on hacking forums reveal that the AI bot can perform various tasks, such as writing code for malware attacks and crafting phishing emails. The creator of WormGPT explicitly described it as a direct competitor and adversary to well-known AI models like ChatGPT, enabling users to engage in illegal activities.

A recent report from Europol, the law enforcement agency, highlighted the exploitation potential of large language models (LLMs) like ChatGPT for cybercriminals. It emphasized that ChatGPT’s ability to generate highly authentic texts based on user prompts makes it a valuable tool for phishing purposes, enabling realistic impersonations of organizations or individuals, even with limited English language skills. Europol warned that LLMs empower hackers to conduct cyberattacks faster, more convincingly, and on a significantly larger scale.

The emergence of WormGPT raises concerns about the misuse of AI tools for nefarious activities, highlighting the need for continued vigilance and enhanced safeguards to prevent cybercrime.

Daniely Kelley, a British computer hacker specializing in cybersecurity, has conducted research on WormGPT, an AI tool designed for cybercrimes like Business Email Compromise (BEC). Kelley’s observations, particularly in underground forums, have shed light on the tool’s capabilities.

Moreover, there is an exchange of specific prompts called “jailbreaks” that allow manipulation of models such as ChatGPT, enabling the generation of output that could potentially reveal sensitive information or execute malicious code.

Kelley’s findings imply a significant implication: even attackers without fluency in a specific language can now fabricate persuasive emails for phishing or BEC attacks with greater proficiency than ever before. These AI-generated emails exhibit grammatical accuracy, which increases their chances of going unnoticed. Additionally, the accessibility and user-friendly nature of AI models have significantly reduced the barriers to entry for carrying out such attacks.

WormGPT is an AI model optimized for online fraud

WormGPT, an AI model specifically developed for criminal and malicious purposes, is being widely shared on popular cybercrime forums. It is positioned as an alternative to legitimate GPT models and marketed with promises of privacy protection and quick financial gains.

In terms of functionality, WormGPT resembles ChatGPT by being capable of generating highly persuasive and strategically sophisticated emails. This makes it a potent tool for executing advanced phishing and Business Email Compromise (BEC) attacks. According to Kelley, WormGPT operates without any ethical boundaries or limitations, distinguishing it from its official counterparts.

WormGPT is built upon the open-source GPT-J model, which offers performance comparable to GPT-3. It can handle various text-related tasks, similar to ChatGPT, and is also proficient in basic code writing and formatting. The precise details of the additional malware datasets used to train the WormGPT derivative remain undisclosed by the tool’s creator.

Kelley conducted an experiment using WormGPT to create a phishing email aimed at deceiving a customer service representative into making an immediate payment for a fake invoice. The email was crafted to appear as if it was sent by the CEO of the specific company being targeted (as shown in the provided screenshot).


By 4niso

Leave a Reply

Your email address will not be published. Required fields are marked *