WormGPT gets a glow-up: crooks swap in Grok and Mixtral for sharper cybercrime
Ourtake
Turns out the bad guys also shop for the best model. Upgrade culture, but make it felony.
What happened
Cato CTRL reported two new WormGPT variants on BreachForums, one reportedly using Grok from xAI via API with a custom jailbreak and another running on Mistral AI's Mixtral. Both are distributed via Telegram and bypass the original models' safeguards to generate phishing emails and malicious code.