A Modern Privacy Lab

Hide sensitive data before you paste into AI

Hide sensitive data before you paste into AI. Restore it after. Nothing leaves your browser.

No uploads. No server. Mapping stays in this tab.

Paste your original text

Default Rules (Ordered by priority)
Optional Extras (Off by default)
Names and custom words to hide

Safe text to paste into AI

Paste the AI's reply here

Restored text

Frequently Asked Questions

Learn how the browser-only privacy and restoration architecture works.

How does it work?

The redactor evaluates your text using an ordered priority pipeline: API keys → URLs with query secrets → emails → card numbers → IP addresses → phone numbers → optional IDs → custom names. Matched entities are swapped with deterministic tokens like [EMAIL_1]. When ChatGPT or Claude replies, the restore parser substitutes original values back in a single pass.

Why is it 100% safe?

Everything runs 100% locally in your browser sandbox using pure JavaScript and Web Workers. There are zero backend servers, zero network calls, zero tracking analytics, and no localStorage of your raw text. You can inspect DevTools Network tab or disconnect Wi-Fi entirely; the tool works completely offline.

What can't it catch?

Implicit contextual clues or informal phrasing without standard patterns (e.g., "my colleague in Munich whose brother founded the startup"). Always inspect the "Safe text" panel and use the "Names and custom words to hide" box to protect proprietary project codenames or person names.

How are AI formatting quirks handled?

Large language models often modify token styling in code blocks, bold tags, or changed cases. The restoration engine is tolerant of variations such as `[EMAIL_1]`, **[ PHONE_1 ]**, [email_1], and escaped brackets \[EMAIL_1\], restoring them accurately while warning if any unknown token appears.