Exposed AI Model Endpoints & Compute Hijacking
Exposed AI Model Endpoints & Compute Hijacking
Articles on unauthenticated AI model servers (Ollama, Dify, vLLM) exposed on the public internet, and how threat actors hijack them for free GPU compute to generate phishing content, malware, and reconnaissance scripts.
exposed ai model endpoints compute hijacking2 hours ago4 min
PoeLLM and AI Driven Endpoint Security Trends Nobody Planned For
Lumen Black Lotus Labs discovered PoeLLM, a malware campaign compromising over 3,400 exposed AI servers for cryptomining and exploit propagation using GitHub-hosted poetry for C2.