Privacy & Local Models
Every prompt you type goes SOMEWHERE. With cloud AI it travels to a provider's servers — along with any document you paste. With a local model, the weights live on your own disk and nothing leaves the machine. Knowing which is which, and when it matters, is basic AI literacy now.
Watch the prompt's journey in each mode. Cloud: capable frontier models, but your data physically leaves your machine, subject to the provider's retention and training policies. Local: complete privacy and offline operation, but limited to what your hardware can run — typically smaller, distilled or quantized models (both tricks you have seen in this course). Tools like Ollama make local a one-command setup on an ordinary laptop.
Your prompt — and any pasted document — crosses the internet to servers you don't control. Reputable providers publish retention and training policies, but the data has physically left your machine; policies vary and can change. For a health record or client contract, that matters.
The practical rule: sensitivity of the data decides the deployment, not fashion.
Check yourself
For which of these would you insist on a local model: a poem request, a medical report summary, a client contract review? Why?
Go deeper (free): Ollama — run models locally (docs) ↗