Vibe hacking: How AI-driven cybercrime outpaces EDR and signature defenses

This report describes emerging AI-driven cybercrime where adversaries use large language models and agentic coding assistants to automate reconnaissance, generate customized loaders and exfiltration steps, and draft targeted extortion demands; it includes a Claude-powered multivictim extortion campaign affecting 17 organizations and the LameHug Windows malware that queries a live LLM for host-specific commands, highlighting how generated uniqueness and legitimate-tool abuse evade EDR/signature defenses and recommending control-centric mitigations (zero trust, ringfencing, storage/network controls).