The detection engine
What Bombus is, how it reaches a result, where detection runs, and what our own test says.
All detection in BeeSensible runs on Bombus. One engine, for every organisation, in the browser, in the desktop app and for documents.
Bombus 2.1
Bombus combines fixed rules for structured data (bank numbers, national IDs, card numbers, email addresses) with small, fast AI models for everything that depends on context (names, addresses, medical and other sensitive details). The layers are merged into one result; overlapping hits become a single marking.
It recognises 65 data types in 11 categories: person, location and address, contact, identifiers, organisation, medical, financial, technical, demographic, travel and time. In Dutch and English. Which of them get marked in your organisation, and how strictly, is up to you through detection profiles; words that only mean something inside your organisation go in as custom terms.
Where Bombus runs
On BeeSensible's own servers at European providers: the application at Scaleway in Amsterdam, the models on a dedicated server at Hetzner in Germany. Both are European companies, not a European region of an American cloud, and both are ISO 27001 certified.
The extension and the desktop app send the text there. It is analysed in working memory and discarded straight after; nothing is kept as content, and no external AI service is involved. What BeeSensible does keep is, per marking, the data type, the level, the app and the moment, without a user id and without the text.
What our own test says
On our own test set, Bombus reaches an F1 score of about 96%, across all 65 data types, in Dutch and English. That is our own test on a frozen test set that was not used for tuning, not an independent measurement and not a certification. The figure says what the detection finds, not how many data breaches are avoided.
Last updated:
