Sensitive documents that mask themselves on entry

The document enters once: the sensitive parts stay masked

When a document lands in the workspace, the sensitive entities inside it become numbered placeholders - the same person, the same account number, one consistent mask across the whole document. The staff keeps reading and asking as usual; the machine never works on the raw values. The only clean copy sits encrypted - the mask is not a trick of the display: the raw text simply is not there to quote.

Only the right people, only on purpose

Need the real number? There is one door: an explicit permission on that workspace (per person), the reading is logged as every operation, and the activity log cannot be altered - you can verify who read what, when. Without the permission the material stays masked - including for machines: the widget and the external tools never receive re-composed text, server-enforced.

An honest empty answer

When the staff asks about a masked field, the reply says the material is masked - it does not guess. Where the documents hold the answer it arrives cited; where they do not, the workspace reads its own configured reply. Nothing gets invented: the census of claims is the standard this platform writes by.

See it live

Richiedi una dimostrazione - bring a folder with the documents you would worry about, and watch the masks land number by number:

• Request a demo
• The operator guide(/en/docs/document-dlp-pipeline)
• The engineering deep dive(/en/blog/dlp-document-pipeline)

Facts in this post - FACT-CENSUS rows F-DLP-1, F-AUDIT-1 (cited in the plan). Provenance: the product docs at v0.28.1.