Anonymising documents before your team shares them
Which files your team can anonymise, how the review screen and the check afterwards work, which profile applies and what you see as an admin.
With the Document Redaction module, your team turns a file into an anonymised copy before anyone shares it or puts it into an AI tool. Think of a quotation with client details that a colleague wants an AI tool to rewrite, or meeting notes with names that are going to an outside party. The copy has the same format as the original, so a Word file comes back as a Word file. When the module is on for your organisation, every member of the organisation can anonymise documents. This article describes how that works, which files are accepted and what you see as an admin.
How a document is anonymised
- Open Documents in the dashboard, or the Documents tab in the extension's sidebar, and pick a file.
- BeeSensible highlights the sensitive data that the detection profile for documents covers, such as names, account numbers, citizen service numbers (BSN), addresses and the organisation's custom terms.
- Every finding is ticked. Untick anything that can stay, draw a box over anything the detection missed, or mask a word throughout the document in one go.
- Check the result with Preview and click Download.
In the copy, the selected text has actually been removed rather than just covered, and properties such as author and company are empty. The original stays unchanged. BeeSensible does not keep drafts, so anyone who closes the screen before downloading starts again.
From the dashboard and the extension, the document goes to BeeSensible's own servers at European providers. Bombus processes it in memory, and BeeSensible deletes it straight afterwards. You choose which profile applies to documents through the Documents target in detection profiles; if you have not assigned one there, the default profile applies.
Which files it accepts
File formats
| Format | What happens |
|---|---|
| Pages without a text layer go through text recognition (OCR). Attachments, scripts and hidden application data are always stripped. | |
| Scans and images (png, jpeg, webp, tiff, bmp, gif, heic) | OCR reads the text; the redaction is drawn into the pixels themselves. Location and device data (EXIF) is always removed. |
| Word, Excel, PowerPoint | Including the old binary formats. BeeSensible also cleans hidden parts: comments, tracked changes, author names, hidden sheets, notes, master slides, alt text, links. |
| OpenDocument (text, spreadsheet, presentation), RTF | The same rules as Office. |
| Plain text, Markdown, CSV, TSV | The encoding is preserved. |
For files that are not PDFs, the review screen shows all the text in the file, hidden parts included, with a heading for each part. This means the user also sees, for example, the comments in a Word file or a hidden sheet in Excel. A file can be at most 25 MB and contain at most 200 pages or 10,000 paragraphs or rows. BeeSensible cannot read password-protected files. It does not recognise faces or signatures; the user draws a box over those themselves.
After every redaction, BeeSensible reads the result back in and checks whether any removed value still appears and whether the file is still intact. If either check fails, the user still gets the file, with the warning "Look at the result before you share it." The user then decides whether to use the file.
The More file formats for documents setting under Policy > Masking is on by default. If you turn it off, your team can only anonymise PDFs in the dashboard and in the extension. The setting itself is described in Masking.
Where your team finds it
| Place | What it offers |
|---|---|
| Dashboard | Documents in the menu. Members without an admin role see only this part. |
| Extension | The Documents tab in the sidebar; drag a file in straight from the page. With Mask on attach, a file someone adds in ChatGPT or Claude comes here first, and the copy then goes into the conversation. |
| Desktop app | The same review screen, PDF only. The user chooses where to save the copy. |
If the document is meant for an AI conversation, the user chooses for each document how removed data appears: as a Black box, where the text disappears, or as a Placeholder such as [NAME1]. With a placeholder, the AI tool can still tell that a name stood in that place, and with Keep context the user sees the real values again in the answer.
What you see as an admin
Insights > Documents shows, for each period, the number of Documents anonymised, Sensitive items found, Removed and Left in on purpose. Below that are the trend over time, the data types, the document types and the way items were removed. The document type is an estimate based on the first page. With department insights on, you also see a breakdown by department.
The dashboard only shows counts. It contains no document content and no per-employee view. If your organisation has fewer than five active users, a card explaining why the figures are not yet visible appears in place of the figures.
Insights > Documents in the dashboardYour organisation's document figures, with period, document type and data type filters.Documents is missing
The Documents section belongs to the Document Redaction module. If that module is off for your organisation, Documents does not appear in the dashboard, the extension or the desktop app, and Mask on attach does not work either.
Last updated:
