Redactic finds the names, tax codes, IBANs, addresses, e-mail addresses, phone numbers and identity numbers in a decision, a case file or an attachment, and your staff decide what each copy shows. Black out the copy for the online notice board, for a requester or for a contractor, and the text under every black box is removed from the file. Mask the same file with placeholders, and your staff can ask an AI to summarise it while the real names come back only in their browser.
Decisions and resolutions, orders, files for access requests, tender documents, letters to citizens, lists of recipients of grants and refunds, and scanned case files, in PDF, Word, Excel or CSV, or as scans and photos.
The free trial masks the first page of one PDF, Word or text file and shows it in your browser, with no account.
Permanently black out data directly on the document. Ideal for securely sharing files with third parties.
Replace sensitive data with placeholders before using AI. Restore the original values in the final output.
Below is an extract from a fictitious municipal decision on a tax refund, as an office would prepare it for publication. Redactic finds the personal data, confirms the Italian tax code and the IBAN by their check characters and leaves the amount, the legal basis and the outcome readable.
| Field in the document | Type of data | How Redactic handles it | Check character verified |
|---|---|---|---|
| Giulia Bianchi | Person name | Recognised; with the list of names, masked with certainty in every form, surname alone included | Not applicable |
| 12/05/1982 | Date of birth | Recognised when introduced by "born on" or "date of birth" | Not applicable |
| BNCGLI82E52F205V | Italian tax code (codice fiscale) | Recognised | Yes |
| Via Solferino 32, 20121 Milano (MI) | Postal address | Recognised | Not applicable |
| IT76W0306909606100000123456 | IBAN | Recognised | Yes |
| giulia.bianchi@example.it | E-mail address | Recognised | Not applicable |
| 18342/2026 | Protocol number | Masked with one click, or with a search that marks every match | Not applicable |
| EUR 1,240.00 | Amount | Kept | Not applicable |
Fictitious data. The amount, the legal basis and the outcome of the decision stay as they are, so the public can read what was decided.
Mask the file and open the document chat: the file reaches the AI with labels where the names, tax codes and IBANs were, and the questions your staff type with real names are masked too before they are sent. Under each message, "What the AI sees" shows the exact text the model received. The answer comes back with the real values restored in the browser, and if masking is not available, the message is not sent.
With Firm the whole office has the Secure Chatbot, an AI chat for questions that start from no document, with the same masking on every message. The people who manage your workspace choose the AI models the team can use, can limit them to providers hosted in the EU or with zero data retention, and decide who uses the chat and how much AI each person can use. The chat keeps no conversation on our servers.
For an organisation with more than 50 users, Enterprise gives each office or department its own AI quota and keeps a log of every operation by user, without any document content.
The signed act stays in your records as it is, and the copy that goes online can carry less. Upload the decision, check what Redactic has found, keep in clear the parts the public needs to read and download a PDF where the rest is gone from the file. Article 5(1)(c) of the GDPR asks that personal data be limited to what is necessary for the purpose, and a copy prepared this way gives the notice board, the transparency section or the requester that and no more.
Add a PDF, a Word file, an Excel or CSV list, a scan or a photo, or a ZIP with a whole case file, which Redactic redacts document by document with the same placeholder for the same person in every file. Choose up to four countries for the document and add the names that must always go to the list, so each one is masked with certainty.
Open Review and edit and read the copy page by page as it will be downloaded. Keep a value in clear with one click, select a word the detection missed, search a protocol number and mark every match at once, and draw a box over a signature, a stamp or a photo. In a list of recipients, mask the columns with names and tax codes and keep the amounts.
Download the blacked-out PDF with its redaction certificate, ready to publish or to send. Or keep the masked version and ask the AI your questions with the names as they appear in the file.
The downloaded file carries no comments, tracked changes, hidden sheets or document properties with the names you removed.
The text under every black box is deleted from the PDF, so it cannot be selected, copied or recovered from the copy you publish. Scanned pages and photos are read with OCR first, and the words found are blacked out on the page.
Old files scanned to PDF and photos of documents are read with OCR, and Redactic blacks out the values it finds. In Review and edit you draw a box over a signature or a stamp, and the "Black out all images" option turns every image in a PDF or Word file into a black box.
Upload a ZIP with up to 200 documents, and Redactic redacts each one with the same placeholder for the same person across the whole archive and returns a new ZIP with the redacted copies.
The redaction certificate records the SHA-256 fingerprint of the original and of the redacted file, and how many items of each type were removed. It contains nothing from the document, so you can keep it in the case file.
Word files lose comments, tracked changes and author details, Excel files lose hidden sheets and cell notes, PDFs lose annotations, attachments, bookmarks, hidden layers and their metadata, and photos lose their location data. Before you download, a final check reads the whole file again and stops the job if a value is still there.
Detection and redaction run in Europe on infrastructure Redactic controls, without third-party AI reading the unredacted text. Your files are not saved, and the data processing agreement under GDPR art. 28 is signed when you subscribe.
Next to each rule is what Redactic does in daily work, so your data protection officer can judge how it fits your organisation.
Personal data in official documents held by a public authority or body may be disclosed under the Union or national law that governs it, to reconcile public access with data protection. When that law asks you to disclose the document and withhold some personal data, Redactic prepares the copy: you choose what stays readable, and the rest is removed from the file.
A public body processes personal data because a task in the public interest or its official authority requires it, and the data must be limited to what that purpose needs. Each copy you prepare in Redactic carries only the data you chose to keep for that purpose.
Article 32 names pseudonymisation among the measures for the security of processing. When your staff use AI, Redactic pseudonymises the file first, and the map that links each placeholder to its real value stays on Redactic's side and expires with the session.
Your organisation decides why and how its documents are processed, and Redactic acts as your processor under a data processing agreement signed when you subscribe.
Redactic describes what its tools do; the assessment of your own processing remains yours, with the support of the DPA and the Trust & Security page. Security
One that removes the text from the file, reads scanned pages and lets staff check every item before the download. Redactic blacks out the values in the PDF itself, reads scans and photos with OCR, shows the copy page by page in Review and edit and adds a redaction certificate with the SHA-256 fingerprints of the original and the copy.
No. The text under each black box is deleted from the exported PDF, and there is no map to bring it back.
Yes. Upload a ZIP with up to 200 documents, and Redactic redacts each one with the same placeholder for the same person across the archive, then returns a new ZIP with the redacted copies for you to review and send.
Yes. Redactic proposes a type for each column whose header it recognises, you confirm or choose the treatment for every column, and the masked columns are covered cell by cell while the amounts and the structure of the sheet stay as they are.
Yes. Mask the file and use the document chat: the AI receives placeholders for the names, tax codes and IBANs Redactic detects, and the real values return in the browser. Under each message you can read the exact text the model received.
Yes. The administrator chooses the models and the default one, and can limit the list to providers hosted in the EU or with zero data retention. Workspaces in the EU data zone see only models hosted in the EU.
No. Masked text is pseudonymised, and it remains personal data under the GDPR for whoever holds the map. For a copy that is published or sent outside, use Blackout, which removes the values from the file, and read the copy once before you publish it, because the context around a value can also point to a person.
Yes. With Firm, up to 50 people have document redaction and the Secure Chatbot, each with their own AI quota. Enterprise goes beyond 50 users, organises them by the units of your organisation, gives each user and each unit a hard or soft AI quota and keeps an operations log per user without any document content.
Detection and redaction run in Europe on infrastructure Redactic controls, and your files are not saved. Enterprise can also run on a dedicated single-tenant server in the EU, provisioned per contract. Your organisation is the controller for its documents, and Redactic is the processor under GDPR art. 28, with a DPA signed when you subscribe. Data Processing Agreement Security
Tell us how many people would use Redactic, which documents they work on and how many each month, and we prepare a proposal for your organisation.
Enterprise plans have no cap on uploads or pages, are invoiced annually and can run on a dedicated server in the EU.