Jailbroken AIs are helping cybercriminals to hone their craft

Researchers observed cybercriminals using jailbroken LLMs (WormGPT variants leveraging Grok and Mixtral) to bypass safety guardrails and generate malicious content—such as malware code, phishing emails, and social-engineering scripts—offered on underground forums; while these tools scale and simplify criminal operations (ransomware and information stealer deployment), they predominantly enhance existing tactics rather than creating entirely new malware families.