WormGPT: The Blueprint for Malicious AI

WormGPT traces the rise of a brand of jailbroken malicious LLMs sold on forums and clear-web sites that enable phishing, keyloggers, ransomware and other malicious code generation; researchers purchased access, demonstrated unsafe outputs, and a February 2026 BreachForums post claims the wormgpt.ai user database (≈19,000 records including emails and payment/subscription data) was leaked — illustrating how these services both facilitate scalable cybercrime and expose their own users to data theft.