Privacy · 2 min read
What "private" and "offline" AI actually mean, and how to check
Many AI apps call themselves private. Here's what actually leaves your computer with a local AI assistant, what doesn't, and how to verify it yourself in five minutes.
"Private" is on every AI product's homepage. It can mean anything from "we don't train on your chats" to "nothing ever leaves your computer". Those are very different promises. Here's how to tell them apart, and how to check one yourself.
Three kinds of "private"
- Policy-private: your messages go to the company's servers, and they promise not to train on them or keep them long. You're trusting the policy.
- Account-private: same as above, with settings to turn history or training off.
- Local: the model runs on your own computer. Your messages are never sent anywhere, so there's no policy to trust.
layman is local by default: answers come from a model running on your PC through Ollama.
What layman sends, and when
| Stays on your computer | Leaves it, only when you choose |
|---|---|
| Your questions and answers, with a local model | Your messages, if you pick a cloud model (labelled in the menu) |
| Selected text, window text, attached files | Web searches you ask for |
| Chat history, settings, snippets, to-dos | Model downloads from Ollama's library |
| Your cloud API keys (encrypted) | A license check with Lemon Squeezy every few days: your key and a device label |
There's no analytics or tracking in the app. layman reads your screen only when you press the shortcut, and never reads password fields or password managers. The full list is on the privacy page.
Check it yourself on Windows
- Open Resource Monitor (type
resmonin the Start menu) and go to the Network tab. - Under TCP Connections, tick
layman.exeandollama.exe. - Select some text, open layman, and ask a few questions with a local model.
- The only remote address you should see is
127.0.0.1: your own computer, where Ollama is running.
On a Mac, run nettop -p layman in Terminal, or use a firewall app like Little Snitch.
Does it work offline?
Yes. After setup has downloaded a model, layman answers with no internet at all. It needs the internet only for web searches, cloud models, model downloads, and a license check every few days. It keeps working for up to 30 days offline between checks.
When a cloud model makes sense
For the hardest problems, a large cloud model is still smarter than anything that fits on a laptop. You can add your own key and use one when you choose, with the provider clearly named. See using ChatGPT or Claude with your own API key.
More guides
- How to run AI locally on Windows (no cloud, no account)A plain-English guide to running a ChatGPT-style AI on your own Windows PC: what you need, which model to start with, and the easiest way to set it up.
- Which local AI model can your PC run? A RAM and GPU guideMatch a local AI model to your computer's memory and graphics card. Sizes for Qwen3, gpt-oss, Gemma, Phi and Mistral, plus a simple rule for what runs fast.
- How to summarize a PDF with AI without uploading itSummarize contracts, reports and Word documents with AI that runs on your own computer, so confidential files never get uploaded to a chatbot.