Skip to content
Walkthrough

Privacy: Redact a Document Before the AI Reads It

A walkthrough of the Privacy app: redact a PDF, Word or CSV file on your own device, check the warnings, download an AI-safe copy, and restore the reply.

Yantra Team3 min read

On this page

The contract you want a second opinion on has names, addresses and account numbers in it. Privacy replaces them on your device, hands you a copy that is safe to share with AI, and puts the real values back into the answer afterwards.

Open it

Pick Privacy in the apps launcher, or open /c/<chat-id>/privacy. A Runs on this device badge sits at the top. The document is read and redacted in the page, and nothing about it is uploaded. Once the page has loaded, it even works offline.

Add a document

Drop a document here, Choose a file, or Try a sample to see how it works first. PDF, Word, CSV and text files are opened by the page itself. For an email or a form export, choose Paste text instead. A scan or a photo has no text to read in the browser; the desktop app can read it.

Redaction starts as soon as the document is in.

Choose what to redact

What to redact is grouped by type, and each toggle shows how many it found:

  • People and places: names and street addresses.
  • Contact details: email addresses and phone numbers.
  • Identifiers: social security numbers, dates of birth, NPI and DEA numbers, and account or long ID numbers.

Under Your terms, add anything else that should never leave: a client name, an employer, a codename. Your terms are redacted wherever they appear, in every document. Privacy also suggests terms it found in the document, such as the parties to a contract, and adds them in one click. Every change runs the copy again.

Check before you share

If anything looks risky, a box lists what to check before you share. That means a value that may have been cut rather than removed, or a pattern no detector claimed. Always redact removes it everywhere and runs again.

Choose how replaced values read. Labels are bracketed tags such as [PERSON_1]. Stand-ins are realistic fake values that read naturally. The Copy tab shows the result, with Show originals to compare. Replacements lists every value and what replaced it, and you can keep a value as written if it is not sensitive.

Then Download AI-safe copy or Copy text.

Restore the answer

Send the copy to any AI you like. When it answers, open Restore reply and paste what it wrote. Every label and stand-in is swapped back for the real value, on your device, and Copy restored text gives you the result.

Two engines

In a browser, the engine is pattern-based. It catches the parties named in a contract, the name and address a letter is mailed to, titled names, anything you add, and every cell of a personal column in a table. It will not catch every stray mention. On the desktop app, names and addresses come from a model running on your machine. The engine used, and any downgrade, is always reported. The desktop panel also shows a privacy audit: details redacted, content blocked from AI, and events.

The assistant and the main chat

Privacy has its own assistant beside the page. It reads only the AI-safe copy, never the original. What you type to it is redacted on your device first. Ask it to "always redact Acme Corp", "use stand-ins instead of labels" or "open the sample".

You do not need the app for every file. The main chat composer has a Privacy switch. With it on, documents you attach are redacted on your device and only the AI-safe copy is sent. The setting is kept on your account.

What it is not

Privacy makes a document safe to share. It does not review it. For a contract review, use Contract Review, which runs the same kind of redaction before anything is sent.

Read more

Keep reading

All posts