Affected Systems
Exposed AI services including LiteLLM, Ollama, Gotenberg PDF converter, Gitea development toolkit, and Ivanti Sentry. Over 3,400 servers compromised globally, primarily in the United States and Western Europe. Active since April 2026 with increasing activity.
Exploitation Status
Active exploitation confirmed. PoeLLM malware actively exploits CVE-2026-42271 (LiteLLM MCP server test endpoints) chained with CVE-2026-48710 for unauthenticated remote code execution. Over 3,400 servers compromised with 11 C2 servers operational. Malware scans ports 3000 and 4000 to identify vulnerable targets.
Business Impact
Compromised servers are weaponized for cryptomining via Kryptex service and converted into scanning/exploit infrastructure for lateral spread. AI/LLM implementations running on high-performance GPU clusters face resource exhaustion, degraded performance, and potential data exposure. Infected systems actively scan and exploit other targets, creating legal and reputational risk. Malware uses novel C2 retrieval via GitHub-hosted poems, complicating detection and blocking.
Urgency
🔴 Immediate
Recommended Actions
- Immediately patch LiteLLM installations to address CVE-2026-42271 and CVE-2026-48710 RCE chain
- Restrict internet exposure of LiteLLM (ports 3000, 4000), Ollama, Gotenberg, and Gitea to trusted IP ranges only; implement VPN or bastion host access
- Search network logs for connections to Black Lotus Labs published IoCs, including Kryptex mining pool traffic and GitHub repository 'dash.css' file requests
- Scan for ELF binary named 'libgcrypt' in unexpected locations and inspect running processes for XMRig or Iron cryptocurrency miners
- Monitor outbound scanning activity on ports 3000 and 4000 from AI/LLM infrastructure as indicator of compromise