Journal

Practical guides on securing AI agents. Every post includes commands you can run today.

2026-11-10 · 6 min read

Is Your AI Agent Running as Root? Here's How to Tell (and Fix It)

privilegessystemdhardening

The single most dangerous misconfiguration on servers running AI agents: the agent process has full root access. Here's the one-liner to check, and the exact systemd changes to fix it.

2026-11-10 · 5 min read

The Kill Switch Your AI Agent Needs (and Why systemctl stop Isn't Enough)

kill-switchincident-responseaiguard

If the only way to stop your agent is to ask it nicely or SSH in and kill the PID, you don't have a kill switch. Here's how to build one that works even when the agent is compromised.

2026-11-10 · 4 min read

Ollama Is Listening on Every Interface. Here's the One-Line Fix.

networkollamaexposure

Default Ollama installs bind to 0.0.0.0, exposing your local LLM to the public internet. Anyone can query your models, see your prompts, and send inference requests. One environment variable fixes it.