Redact personal data, keep the document useful.
Redactic is a European document redaction tool. It finds personal data in PDF, Word, Excel, images and scanned files, and lets you either mask it reversibly before working with AI or black it out permanently before sharing. Files are processed in the EU, held only in encrypted memory for a few minutes, and never used to train any model.
- Processed in the EU
- No document retained
- Never used for AI training
- Restoration in your browser
- DPA signed on subscription
Two ways to protect a document
Pseudonymization
Names, tax codes, IBANs and addresses are replaced with placeholders. You can run an AI task on the masked text, inside Redactic or with your own model, and the original values are restored locally in your browser. The mapping never reaches the server.
Anonymization
Sensitive data is blacked out directly on the page, including scanned PDFs and images read with OCR. There is no map and no way back, so the downloaded file carries nothing beneath the black bars. Hidden metadata and author properties are stripped as well.
Pseudonymization or anonymization: what GDPR expects
The two terms are often used as synonyms, but the GDPR treats them very differently. Pseudonymized data, as defined in Article 4(5), is still personal data: the identifiers are replaced, yet the information can be linked back to a person with a separate key. It stays within the scope of the regulation, although Article 32 lists pseudonymization among the measures that lower risk. Anonymized data can no longer identify anyone, and Recital 26 places it outside the GDPR altogether.
The right choice depends on what happens to the document next. When you need the original values back, for example to finish a contract draft or a payroll report that an AI model helped you write, pseudonymization keeps the work usable. When the document leaves your organisation, such as a court filing, a response to an access request or a file shared with a supplier, permanent anonymization removes the data for good. Redactic offers both modes from the same upload, so the decision can be made document by document.
Redacting documents before using ChatGPT and other AI tools
Pasting a contract or an employee file into a chatbot sends every name, tax code and bank detail to the model provider. Redactic sits in front of that step: it masks the personal data, lets you run the task on the masked text inside Redactic or with the AI tool you already use, and restores the real values in your browser once the answer comes back. The model only ever sees placeholders, and the mapping between placeholders and real values never reaches our server.
Masking, permanent blackout and redaction certificates do not consume AI credits, so the protective step costs nothing extra on any plan. Credits are used only when you run AI tasks inside Redactic.
Supported files
| Format | What Redactic does |
|---|---|
| PDF (text and scanned) | Detects personal data, including OCR on scanned pages, and blacks it out on the page or masks it for AI |
| Word | Masks or removes personal data in the text and strips author properties and hidden metadata |
| Excel and CSV | Redacts by column while keeping the table structure intact |
| Images | Reads text with OCR and blacks out the detected values |
| Plain text | Paste and mask directly, without uploading a file |
How it works
- Upload or paste. A PDF, Word file, spreadsheet, image or plain text. Spreadsheets can be redacted by column while the table structure is preserved.
- Review what was detected. Machine learning models and deterministic rules propose the entities, and you decide what to mask or keep.
- Share or run AI. Download the redacted file, or send only the masked text to a model. A redaction certificate with document hashes and a summary of removed data is available for every operation.
National identifiers validated by check digit
Generic detectors recognise names and emails well but often miss the identifiers that carry the most risk in European paperwork. Redactic recognises the country formats below and validates them by check digit, so that look-alike numbers such as invoice references are not mistaken for real identifiers.
| Country | Identifiers |
|---|---|
| France | NIR, SIREN, SIRET, numéro fiscal, RIB, TVA, passport |
| Germany | Steuer-IdNr, USt-IdNr, Sozialversicherungsnummer |
| Italy | Codice Fiscale, Partita IVA, Tessera Sanitaria |
| Spain | DNI, NIE, NSS |
| All countries | IBAN in any country format (mod-97), payment card numbers |
Formats for the Netherlands, Switzerland, Austria, Belgium, Ireland, the United Kingdom, the United States, Canada, India, Singapore and South Africa are also recognised, together with names, companies, postal addresses, email and certified email (PEC), phone numbers, passports, identity cards and driving licences. The full and current list is on the Redactic website.
Who uses it
Redactic is designed for professionals who handle personal data every day and need a quick, verifiable way to remove it: law firms and their assistants, HR and payroll offices, accountants, labour consultants, notaries and administrative staff.
Questions
Are my documents saved?
No. The document and the result stay only in encrypted memory for a few minutes and are then deleted. There is no database copy, no log of the content and nothing written to disk.
Do I have to use your AI?
No. You can run the task inside Redactic or copy the masked text and use it with any model you prefer. In both cases the real values come back only in your browser.
Who is the data controller?
You are. Redactic acts as the data processor and the data processing agreement is signed automatically when you subscribe. The legal basis and the notice to the people in your documents remain with you.
Does it work with scans?
Yes. Text in scanned PDFs and images is recognised with OCR, masked and, if you only need to share, blacked out directly on the page.
Is pseudonymized data still personal data under GDPR?
Yes. Under Article 4(5) GDPR pseudonymized data remains personal data, because it can be linked back to a person with additional information. Only anonymized data, which can no longer identify anyone, falls outside the GDPR as described in Recital 26.
Further reading
- Redaction Notes: short, sourced guides on PDF redaction failures, tenant files in France and Germany, and placeholders before AI
- Regex is not enough: validating EU national IDs and IBANs by check digit in Python (DEV Community)
- Redactic for law firms and for real estate agencies
Listed on
Redactic is listed in these startup, software and legal tech directories, where you can read the product profile, follow updates and leave a review.
- Redactic on Product Hunt
- Redactic on LaunchIgniter
- Redactic on Startup Fame
- Redactic on The One Startup
- Redactic on StartupTrusted
- Redactic on SaaSGrow
- Redactic on Trustiner
- Redactic on Twelve Tools
- Redactic on F6S
- Redactic on Mapa LegalTech Polska
- Redactic in the Village de la Justice LegalTech guide