Cryptomining Botnet Uses GitHub Poem to Hide C2, Compromises 3,400 AI/LLM Servers
What Happened – A new cryptomining botnet, dubbed PoeLLM, has been observed pulling its command‑and‑control (C2) address from a poem stored in a public GitHub repository. The malware exploits known vulnerabilities in open‑source AI/LLM services (e.g., LiteLLM CVE‑2026‑42271) and other self‑hosted tools such as Gotenberg and Gitea, hijacking more than 3,400 servers across the US and Western Europe since April 2026.
Why It Matters for Trust & Control Assurance
- Demonstrates how unmonitored third‑party code (GitHub) can become a covert C2 channel, highlighting the need for continuous monitoring of external repositories as part of a control‑assurance program.
- Exploits publicly disclosed CVEs, underscoring the importance of timely vulnerability remediation and secure configuration of internet‑exposed services.
- The campaign’s scale shows that a single mis‑configuration can translate into a systemic resource‑drain, a scenario that continuous evidence collection and control mapping can detect and prove mitigation.
Who Is Affected – Cloud‑native SaaS providers, AI/LLM platform operators, open‑source tool maintainers, and any organization running self‑hosted AI services.
Recommended Actions
- Inventory all internet‑exposed AI/LLM endpoints and verify they are patched against CVE‑2026‑42271, CVE‑2026‑10520, and similar disclosures.
- Implement continuous monitoring of public code repositories for anomalous files or patterns that could be used for C2 signaling.
- Map the vulnerability‑remediation and external‑repository‑monitoring controls to your audit framework and collect defensible evidence. Source: https://www.helpnetsecurity.com/2026/10/08/poellm-malware-github-poem-ai-servers/
Technical Notes – The botnet scans ports 3000/4000 (default for Gotenberg and LiteLLM), delivers a crafted POST request to trigger a command‑injection exploit (CVE‑2026‑42271). The C2 address is encoded as four words in a poem (dash.css) and decoded at runtime. Initial detection stemmed from an Ivanti Sentry breach (CVE‑2026‑10520). Source: https://www.helpnetsecurity.com/2026/10/08/poellm-malware-github-poem-ai-servers/