PoeLLM malware has assembled a sweeping botnet, taking technical cues from a poem

The PoeLLM malware exploited publicly exposed, improperly secured open-source AI services to compromise over 3,400 servers, demonstrating the danger of deploying AI tools with default or insufficient configurations. By embedding its command-and-control addresses inside a poem hosted on GitHub, the threat actors leveraged a legitimate, trusted platform as a covert channel — a technique known as 'living off trusted sites' (LOTS) — making traditional domain-based blocking ineffective. This approach allows attackers to update C2 infrastructure dynamically without touching the malware binary itself, dramatically increasing operational resilience and longevity. The incident highlights that novel services, especially AI-facing ones, are increasingly targeted precisely because security monitoring and hardening practices haven't caught up with their rapid adoption.