Redact Assisted is one tool for the whole job – from the AI that finds the personal data to the file that leaves the system without it. Here is every feature, as it actually exists in the product.
Language models trained specifically for redaction read the whole document in context. Not keywords, not word lists – an understanding of what is a person, an address or a case number.
Categories recognised out of the box
Missing a category? Create it yourself with a name, alias, colour and your own search pattern (regex). Use it for industry-specific IDs, internal case numbers or words that must always go – side by side with the AI’s suggestions.
“Emma Walker”, “E. Walker” and “Walker” are automatically resolved to the same identity and get the same alias throughout the document.
Four variants in the text → one identity → one alias
Every person, organisation and address is re-searched across the whole text – including misspellings, hyphens across line breaks and missing accents.
Every finding carries a confidence score. Uncertain suggestions are clearly flagged for review, so nothing slips through unnoticed.
Per-category whitelists keep e.g. “City of London” or PO box addresses. Forced patterns make sure specific formats are always redacted.
Choose how people are replaced: a running counter, letters, doubled letters or fixed text – per category, or for groups of categories that should share one numbering.
The standard model covers most needs. If you have a special use case, we train a custom model that understands your terminology and document types – and controls precisely what is redacted and what is kept.
Detection is trained on almost every European language – with Danish, Swedish, Norwegian and English at the sharp end. OCR for scanned documents runs in seven languages.
The AI has done the work; you approve. AnteRedact lists every finding grouped by category next to the document, so you can confirm, adjust and approve in minutes.
Meeting minutes
Client meeting 25 · 12-04-2025
Attendees: Emma Walker, Lucas Moreau, Sofia Rossi.
Emma Walker gave a status update on the schedule. The foundation is complete and the shell starts in week 17.
Lucas Moreau will send revised drawings to sofia@rossi-architects.com by Friday.
Questions about the minutes: tel. +44 20 7946 0123.
Select text in the document, or search – whole words, case sensitivity and hyphenation included. Add all matches to an existing person or a new category in one click.
Drag two identities together when the AI has split one person in two – or split when two people have become one. The aliases follow automatically.
Draw a box over signatures, stamps and graphics. Redact directly inside embedded images. Mark whole pages for deletion.
Word and RTF documents are reviewed in their own page layout – with styles, numbering, headers and footnotes – and redacted in the original format.
See the finished result with boxes and aliases while everything is still editable. Approve only when it looks right.
Unlimited undo/redo, automatic saving every 10 seconds and keyboard shortcuts for search, navigation and approval.
Upload what you have. Word and RTF are redacted directly in their original format, PDFs are processed with their format preserved, and everything else is converted to PDF. Scanned pages and images are OCR-processed automatically.
.pdfTrue content removal – not a box on top
.docxWord in, Word out. Headers, footnotes and properties included
.rtfRedacted directly in the RTF file
.docConverted and redacted as PDF
.pptx.pptAll slides, notes included
.xlsx.xlsSpreadsheets containing personal data
.msgE-mails including HTML body and PDF attachments
.emlThe standard format from every mail client
.jpg.pngScanned letters and photos of documents
.txt.md.html.xml.json.yamlStructure and format preserved
AnteRedact reads the text directly wherever it exists and OCR-processes only the pages that lack it: scanned pages, images and broken text layers. That makes it very fast – and quality never drops more than it has to.
A black box over the text is not redaction. AnteRedact removes the characters from the file itself – in every layer where they could hide – so the result can be shared without risk.
The characters are removed from the file itself and replaced by a box and an alias. Nothing left to copy or search for.
Form fields, annotations and hidden accessibility text are blanked as well.
Vector paths over a redaction are removed. In images, the redaction is burned into the pixels.
Upload hundreds of documents at once, follow their status in real time, approve and download together. Every document keeps its own log of what happened to it.
Drop many files at once. Approve, download or delete documents together. Download the whole batch as a single zip.
Queued → Processing → In review → Redacted. The list refreshes itself, and failures are explained right on the row.
Every processing step is logged with timestamp and details, so you can always document what happened to a file.
Redacted files and drafts are deleted automatically after a number of days you decide. The original never stays longer than necessary.
Every document is checked against your thresholds. If it meets all of them it is redacted automatically – otherwise it lands in manual review. Or force everything to review, if you prefer.
A project holds a case’s documents, its people and its rules. Categories, aliases, thresholds and workflows can differ from the organisation default – per project.
Nobody wants to move files by hand. AnteRedact comes to the documents – on the shared drive, in the document management system or through your own code.
A small Windows app watches the folders you choose and redacts new and changed files continuously – without anyone having to remember.
Server-side connectors that watch and write back – with no client on the desktop.
Three calls: upload, status, download. API keys are created in settings and the documentation lives inside the product.
# Upload a document
curl -X POST $ANTEREDACT_API/documents \
-H "Authorization: Bearer $API_KEY" \
-F "file=@case-3079.pdf" \
-F "review=auto"
→ { "id": "9f3c1a7e", "state": "queued" }
# Poll the status – or receive a callback
curl $ANTEREDACT_API/documents/9f3c1a7e \
-H "Authorization: Bearer $API_KEY"
→ { "id": "9f3c1a7e", "state": "redacted" }
# Fetch the redacted file
curl -o case-3079_redacted.pdf $ANTEREDACT_API/documents/9f3c1a7e/result \
-H "Authorization: Bearer $API_KEY"Other systems – e.g. SharePoint or your case management system – connect through the API. We help with the setup.
AnteRedact handles the most sensitive material you have. That is why everything – including the AI models – runs with us in the EU or with you. No data is sent to third-party AI.
Hosted in Paris with a European provider. Every customer gets its own isolated environment with its own database and storage – not just a row in a shared table.
The same solution installed in your own Kubernetes environment. No data processing agreement needed – data never leaves the building.
All models run in our own infrastructure. No OpenAI, no third-party APIs.
TLS in transit, encryption at rest and optional client-side encryption where only you hold the key.
E-mail verification, TOTP app and recovery codes. Sessions are enforced centrally on every single request.
Invite colleagues or let your whole e-mail domain join automatically. Owners and members per project.
Automatic clean-up deletes files after your deadline. Documents that only pass through – e.g. from Redact Folders – are deleted as soon as the result has been delivered.
Continuous operations monitoring and automatic vulnerability scanning of every component before each release.
The default setup works from day one. If you want to fine-tune, it is all in settings – for the whole organisation or per project.
Book a demo where we redact your typical documents live – or try it yourself for free on redactpdf.io.