Forest Global Training Partners

Cyber Risks · Unit 42 Palo Alto Networks

The Dual-Use Dilemma of AI: Malicious LLMs

Summary: Unit 42 examines two purpose-built malicious LLMs, WormGPT and KawaiiGPT, which remove ethical constraints and safety filters during training or fine-tuning and are marketed in underground forums and Telegram channels. WormGPT emerged in July 2023 reportedly built on the GPT-J 6B open-source model and fine-tuned on malware code, exploit write-ups and phishing templates, while KawaiiGPT was first identified in July 2025, is currently at version 2.5, and is freely available on GitHub with setup often taking under five minutes on most Linux systems.

Why it matters: WormGPT 4 sells tiered subscriptions and full source code, with sales campaigns starting around Sept. 27, 2025 and a Telegram subscriber count over 500, while free tools like KawaiiGPT remove cost as a barrier entirely. In Unit 42's testing, WormGPT 4 instantly produced a functional PowerShell script to encrypt and lock PDF files on a Windows host and drafted a ransom note with a 72-hour deadline after which the price doubles.

Read the original at Unit 42 Palo Alto Networks ↗

Topics
AI Security
Related