self-ai-infra

Self-hosted AI inference infrastructure: local LLM runtimes, serving stacks, and their security surfaces.

research

Six Runtimes, One Pattern: The 2026 CVE Wave Across the Self-Hosted AI Inference Stack

Six runtime families, one pattern: the same five architectural flaws across Ollama, llama.cpp, vLLM, NVIDIA Triton, MLflow, and MCP — an evergreen operator deep-dive of the 2026 CVE wave, with per-flaw detection commands and a durable 12-item checklist.