Build an AI Assistant That Sends No Data to the Cloud — Ever

Docker Model Runner runs the LLM inference entirely on your local machine — you can verify this with network monitoring tools that show no outbound connections to AI provider APIs during inference. OpenClaw is open source so you can audit the code to verify what network calls it makes. The instructor covers verification techniques during the workshop.

For AI inference: none. Docker Model Runner processes messages locally with no cloud API calls. The only external network usage is for WhatsApp and Telegram message delivery (same as any messaging app), Docker Hub for initial model downloads, and Google Fonts for typography (which can be removed if needed). Zero conversation content goes to any cloud AI service.

Yes. WhatsApp and Telegram handle message delivery through their own servers — this is unavoidable when using these platforms. However, the AI processing of those messages happens entirely locally on your machine. The distinction is: your messages travel through WhatsApp's servers as encrypted messages, but the AI inference that generates responses happens only on your hardware.

OpenClaw is configured to use the Docker Model Runner local endpoint as its AI backend. There is no accidental cloud AI fallback — if Docker Model Runner is not running, OpenClaw returns an error rather than falling back to a cloud API. The instructor covers this architecture and how to verify it during the workshop.

Many developers use this setup specifically for work conversations involving confidential information. Because no data goes to cloud AI services, you maintain the same data security standards as keeping information on your corporate network. The instructor discusses workplace deployment considerations during the workshop.

A no-cloud-data AI assistant where processing happens entirely on hardware you control simplifies GDPR compliance significantly — there is no third-party data processor involved in the AI inference. For formal GDPR compliance assessment, consult a legal professional. The instructor can discuss the general data flow during the workshop.