Strip names, addresses
and account numbers
from PDFs
right in your browser

docscrubber automatically detects personal information in PDFs and redacts it for you. Your sensitive documents stay private, on your device.

docscrubber.app / your-document.pdf
Network: 0 B sent
Source document
invoice_aug.pdf · 1 page

From the desk of Eleanor M. Whitfield, Whitfield & Co. Consulting

Billed to Marcus Tan at 418 Linden Avenue, Apt 3B, Brooklyn NY 11215. Please remit payment to account ending •• 4429 by August 31.

Questions? Call (415) 555-0148 or email e.whitfield@whitfield-co.com. Thank you for your business.

Detections found
Scanning…
demo
Files uploaded
0bytes
Detection model
openai/privacy-filter
Runs entirely in
your browser
No documents leave
your device
— 01 / Who it’s for

For anyone who needs to share a document. But not everything in it.

— Individuals

Send your lease, not your social security number.

Sharing a bank statement to verify income? A lease for a roommate application? Black out what doesn’t need to travel with it.

Account holder: Jordan Liu
Routing: 021 000 089 · Acct: •• 7102
Statement period: Jul 1–31, 2026
— Small businesses

Share invoices and contracts without leaking client data.

Sending a sample contract to a new lead? Posting a redacted invoice to your accountant’s portal? Review the document, nothing slips through.

Contract between Northgate LLC and
Ada Pereira · ada@studio-pereira.com
Total: $12,400 · Net 30
— Job seekers & recruiters

Pass a resume around without sharing a home address.

Getting feedback on your CV? Forwarding a candidate to the hiring panel? Strip out home addresses, personal phone numbers, and references in seconds.

M. Aldridge · Senior Product Designer
Address: 148 Elm St, Oakland CA
Email: m.aldridge@gmail.com
— 02 / How it works

Your document goes nowhere.
It stays on your machine.

The usual way cloud tool

  • 1You upload your PDF to their server.
  • 2It’s stored — possibly logged, possibly cached.
  • 3It’s sent to a third-party model API.
  • 4Redacted file is sent back. The original went to the cloud.

The docscrubber way local-first

  • 1You drop your PDF into the page. It stays in browser memory.
  • 2The detection model is downloaded once.
  • 3PII is found and redacted by code running on your machine.
  • 4You save the cleaned PDF. Nothing left your machine.
— 03 / Accuracy

Automatic detection powered by OpenAI’s privacy-filter model.

96
% F1
token-level on PII-Masking-300k
// precision & recall on two benchmarks
PII-Masking-300kMultilingual PII detection
Precision
94.0%
Recall
98.0%
CredDataSecrets in codebases
Precision
74.7%
Recall
96.5%

OpenAI’s privacy-filter is a small, purpose-built machine learning model that was trained to recognize the messy ways personal information actually appears in real documents.

You can see every detection on the page before anything is saved. You can review exactly what will be blacked out. If something looks wrong, you can deselect it. You control the final outcome.

!
No model is perfect.

Even a world-leading model can miss a name in an unusual format or a partial address. That’s why docscrubber always shows you every detection before you save. You review the detections and only download when you’re satisfied.

Get early access to docscrubber.

docscrubber is in early access. Join the waitlist and we’ll let you know when your account is ready.