Google Uncovers Malware Using LLMs to Operate and Evade Detection

Google's threat intelligence analysts have observed AI-powered malware being deployed in the wild, marking what they call "a new operational phase of AI abuse." The report highlights PromptLock, the AI-powered proof-of-concept ransomware developed by NYU Tandon researchers — with doctoral candidate Md Raz lead author on the paper — that was initially mistaken for an active threat by security firm ESET. Google's findings confirm that attackers are now creating similar malware leveraging LLMs to operate autonomously and evade security systems, representing a significant evolution in cyber threats.