Build Your Own Secure Local AI Assistant for Cyber Security (Ollama + Open WebUI)
Channel: Joshua Clarke Format: YouTube video Published: 2025-04-03 Sponsor: None disclosed (Proton VPN/Pass affiliate links in description)
Summary
Joshua Clarke walks through installing Ollama and Open WebUI on Windows to build a private LLM assistant for cybersecurity communication tasks — drafting incident reports, explaining technical findings to non-technical stakeholders, softening tone in emails. The motivating constraint: cybersec data is sensitive and can’t go to cloud LLMs without explicit org whitelisting.
Key Points
- Privacy first — for cybersec workflows, never send sensitive data to a cloud LLM unless your org has specifically whitelisted that service
- Local stack: Ollama for model serving, Open WebUI for chat interface, Docker to run Open WebUI
- Bonus install tip: change Ollama’s default model storage path via Windows environment variables — the default lives in AppData and can fill the C: drive fast
- Model choice: pick from ollama.com/models; capability tags (
tools,embedding,vision) signal which features the model supports - Use case demo: rewriting a curt “you have TeamViewer, get rid of it” email into a structured, professional message via natural-language prompting
- Validation discipline — always review LLM output for accuracy and org-policy alignment; local models still hallucinate
Sponsorship & Bias Notes
Sponsor: None disclosed. Proton VPN and Proton Pass affiliate links appear in description, unrelated to the video content.
Product placement / affiliations: None observed in the video itself. Ollama and Open WebUI are open-source and presented evenhandedly.
Comparison bias: None observed.
Connected Pages
- ollama — model runner
- open-webui — UI layer
- open-source-model-integration — broader cost-and-privacy thesis
- joshua-clarke — author hub
See Also
- summary-nate-herk-ollama-claude-code — different angle: using Ollama as a Claude Code substitute for cost