New NadMesh Botnet Hunts Exposed AI Services for Cloud Keys and Kubernetes Tokens

A newly discovered Go-based botnet named NadMesh has emerged, actively scanning for misconfigured AI and DevOps services to extract sensitive credentials like AWS keys and Kubernetes tokens. The threat actor’s dashboard reportedly lists over 3,800 unique AWS keys harvested from vulnerable systems. QiAnXin's XLab identified the malware based on a string reference to 'n4d mesh controller' in its codebase.

The botnet focuses on popular open-source tools such as ComfyUI, Ollama, n8n, Open WebUI, Langflow, Gradio, Jenkins, Docker API, and Redis—services often deployed without proper firewall protections. It exploits weak configurations, default credentials, and unauthenticated endpoints to gain access. Once inside, it collects cloud credentials stored in environment files, Kubernetes service accounts, and Docker configuration files.

Researchers note that while the botnet uses multiple vectors—including Docker remote code execution and Jenkins script console exploitation—the largest portion of its activity involves exploiting MCP (Model Context Protocol) servers using the execute_command tool. However, many of these services lack robust authentication mechanisms, making them easy targets.

Organizations are urged to secure exposed services by enabling authentication, restricting public access, and applying available patches for known vulnerabilities like CVE-2026-39987 and CVE-2026-41176.