# Security and privacy

- All classification, training, metrics and exports run in your browser as JavaScript.
- The application bundles synthetic research examples only, not private files.
- User-entered examples are processed locally and are not intentionally transmitted over a network.
- Experiment history contains only date, dataset split, seed, model summary and numerical aggregate metrics. It is stored in the browser's `localStorage` (up to 15 entries), and can be cleared from the app.
- No credentials, cookies, telemetry, tokens, external scripts or API endpoints are required by the app.
- Inspect scripts before customizing or rehosting. Do not publish real private research texts as static assets.
- CSV exports are intended as diagnostic artifacts; avoid importing untrusted formula-bearing strings into spreadsheets without checking them.