Lion Cipher Lab
Install Ollama and run your first local model — safely
- AI Security
- Beginner
- ~60 min
- Free · no account required
Why it matters
Almost fifty labs on this site — phishing triage with a local model, log summarization, building a small offline security agent — start with the words "with Ollama installed and a model pulled." This is the lab that makes those words true. It also teaches, on the first day, the single most common way people hurt themselves with local AI: they hear "it runs on my machine, so it's private and safe," and stop thinking. A local model is private by default only because it listens on localhost; it has no password, it obeys injected instructions exactly like a cloud model, and a model file from a…
What you'll learn
- installing Ollama
- choosing a model that fits your RAM
- ollama pull and ollama run
- calling a local HTTP API with curl
- localhost vs 0.0.0.0 binding
- reading an install script before running it
- local-AI threat model basics
- setting up a lab environment
Steps in this lab
- What a "local model" is, and what Ollama actually does
- Hardware honesty: what your machine can actually run
- Install on Windows
- Install on macOS and Linux — and how to read a script before you pipe it to sh
- The server, the port, and where your models live on disk
- Pull one small model that fits 8 GB
- Run it: your first prompt, and how to leave
- Talk to it over the API: one curl call, and what "stream": false does
- The security section: what "local" does and does not protect you from
- Troubleshooting the four things that go wrong
- Prove it: the three-line check every later lab assumes
- What you just built — and the instinct to keep