Free document anonymization tool — process files locally in your browser

Sensitive entities that are detected and confirmed — names, ID numbers, phone numbers, emails, bank/USCC numbers and organizations (via regex plus an on-device NER model) — are replaced with stable placeholders such as ⟦NAME_1⟧. This is also known as document anonymization or document anonymisation. Text that is not detected as a sensitive entity is left unchanged and is not protected, so review the result before sharing.

No. Parsing, detection and anonymization all run in your browser (and a Web Worker). Your document and its sensitive information never leave your device. The page can fetch static assets from a CDN — the AI recognition model, and fonts when you anonymize a PDF. Those asset requests contain none of your content.

No. This tool produces a one-way anonymized copy for sharing — it does not store a mapping table and cannot reverse the placeholders. Keep your original file if you need the real values back.

This page only anonymizes and does not translate, so no document text is sent to a translation provider. The translation flow has two separate paths: BYOK sends masked text and your key directly to your chosen provider, while managed mode relays masked text through Veil Translate to Cloudflare Workers AI. Undetected sensitive data can remain in the body text.

Yes: DOCX main-body text, XLSX cell text, PPTX slide text and speaker notes, and text-based PDFs are supported. Word header and footer text, PowerPoint chart or SmartArt text, scanned PDFs, encrypted PDFs, and rotated or vertical PDF text are not processed. Office file structure is preserved; PDF layout is closely preserved rather than guaranteed identical.

No. Your document content is processed locally in the browser and is never uploaded, but the page downloads static assets such as the on-device recognition model and PDF fonts from a CDN. Those requests contain none of your document content.