WormGPT Is ChatGPT With ‘No Ethical Boundaries or Limitations’
A hacker has created WormGPT, a malicious version of ChatGPT, designed to assist cybercriminals. The developer is selling access to this chatbot in a hacking forum. Unlike ChatGPT, WormGPT lacks any guardrails to prevent it from responding to malicious requests. The developer boasts that WormGPT can be used for illegal activities, offering access to various malicious activities from the comfort of one’s home. The chatbot can generate malware written in Python and provide tips for crafting malicious attacks.
WormGPT was trained on data related to malware creation using an older open-source language model called GPT-J from 2021. SlashNext, an email security provider, tested WormGPT and found it capable of producing convincing and strategically cunning phishing emails, such as business email compromise (BEC) schemes. The chatbot crafted emails using professional language, making them difficult to detect as phishing attempts.
Despite one buyer’s complaint about weak performance, the sale of WormGPT for 60 Euros per month or 550 Euros per year indicates a concerning trend in the potential misuse of generative AI technologies for cybercrime. This experiment underscores the significant threat posed by such AI programs, even in the hands of inexperienced cybercriminals, highlighting the potential dangers as these programs become more advanced.
The title of this article is misleading, and seems insidiously intentionally so. The issue is not that WormGPT has ‘No Ethical Boundaries or Limitations’, it’s that it was literally trained to write malware. The title attempts to conflate preventing malware with the so-called ethical limitations of ChatGPT, which are less about malware and more about pushing The Narrative,. i.e. restricting answers to only align with what the tech overlords have deemed is The Truth. Which as we all know, doesn’t always line up with what is the truth in reality, such as with the covid lab leak.