Clean a document before you share it.
Open a PDF, review what BeeSensible found, and redact it for good. Then share it or hand it to AI for analysis, with the names, IBANs and BSNs gone.
Sensitive data is found automatically
Open a document and Bombus detects names, addresses, IBANs, BSNs and 65 data types without any setup. Everything found appears in the list on the right.
You decide what gets removed
Every item is selected by default. Uncheck anything that should stay in. When you confirm, only the selected items are permanently redacted from the file.
Draw a box yourself
Logo, signature, or something the engine missed? Draw a box over it by hand and that area is masked too. Manual and automatic detections are redacted together.
The method
The same four steps, now for documents.
Every BeeSensible module follows this cycle. Here is what it looks like for the documents your organisation shares.
What leaves your documents
The insights show how many documents were anonymised, which kinds of documents they were, which data types were found in them, and how much of that was removed or left in on purpose.
Which data has to go
The detection profile decides what a document is searched for. While anonymising, you choose yourself what stays.
Clean before you share
Open a PDF, review what was found, draw a box over anything detection missed, and export a clean copy. Hidden document properties are stripped too.
What you can show for it
Anonymised documents, data found and removed, and what people chose to keep, over a period, as substantiation of data minimisation.
and again How the cycle works
Setting what a document is searched for
A profile sets which data types are found and how heavily each one counts. Your organisation makes one of them the default, and when you open a document you pick another profile from the toolbar if that fits better. A file going out of the door asks for more than an internal note.
What the profile finds is a proposal. The screen shows everything found, grouped by kind, and you untick whatever should stay.
Mask a term, or a box that returns on every page
Click a word in the document and you choose whether only that one occurrence is masked, or every occurrence in the whole file. A project name or a brand that is nowhere recognised as personal data is gone everywhere in a single move. If you would rather not use the mouse, you type the word into the panel.
For anything that is not text you draw a box: a logo, a signature, a stamp. If it sits in the same place on every page, one click carries the box through the whole document, instead of drawing it twenty-three times over.
Bombus, our best engine yet.
Bombus 2.1 is the engine behind BeeSensible. A smart combination of fast AI techniques and hard rules, working as one judgement: sharp on structured data like IBANs and medical codes, and just as good on names and addresses that depend on context. On our internal benchmark it scores roughly 96% F1, across all 65 data types, in Dutch and English.
How it works
The safe way to let AI read a sensitive document.
AI is good at summarising and analysing documents, but a contract or a patient file should not go in as-is. Redact it first, then upload the clean copy. The original details never reach the AI tool.
Review first, remove after
BeeSensible finds names, IBANs, BSNs and dozens of other types across 65 data types, and groups them by category. Check the list, uncheck anything that should stay, and draw a box by hand over anything it missed.
Permanently removed
Redaction is permanent. The values are removed from the file itself, and hidden document properties are stripped too, so nothing leaks through the metadata.
The file itself is not kept
The document is processed in working memory on our own servers at European providers and removed straight after. Only the number of items you redacted is counted, so you can safely hand the clean copy to an AI tool.
For the documents you share
Contracts and legal
Take names, case numbers and addresses out of a contract before it goes to a counterparty, or before a chatbot summarises it.
Patient and care files
Strip patient names, BSNs and medical details so a file can be analysed without exposing the person behind it.
HR and finance
Salary data, IBANs and employee IDs removed before a document leaves the team or gets pasted into an AI tool.
See what came out of your documents.
The insights add up what the module delivered: how many documents were anonymised and of which kind, how much sensitive data was found in them, how much of it was removed, and how much a person chose to leave in. Per data type you see found next to removed, and you see whether things went out through the detections or through a box someone drew or a word they clicked. Counts only, aggregated per organisation, never per person and never a value. Switch department insights on and you see it per department too, with a floor so a single person can never be read from it.
Try it on a document of your own.
Open a PDF in the desktop app or drag it into the extension's side panel, review what is found, and produce a clean copy you can share.
