Ga naar hoofdinhoud
Solutions/OpenAnonymiser
BetaSector: Public & SMB

OpenAnonymiser. Strip PII before it ships.

A Dutch-first PII detection service for Nextcloud. OpenAnonymiser reads any text or document, finds persons, addresses, IBANs, BSNs, and case numbers, then hands DocuDesk the spans to redact. Ships as a Nextcloud ExApp with a CPU-only build (Light) and a GPU build. Presidio under the hood, EUPL-1.2.

The outcome

Nothing personal leaves without your knowledge.

Install the ExApp, point your DocuDesk and OpenRegister flows at it, and every document and record gets scanned for PII before it publishes, exports, or syncs out. The findings come back with explainable spans you can review or auto-redact.

Dutch-trained out of the box.

Two builds for two budgets.

Wired to DocuDesk and OpenRegister.

Explainable, auditable, EUPL-1.2.

Async by default.

Two minutes to install.

The pipeline

From any source to a clean publication.

OpenAnonymiser sits as a sidecar to your existing Conduction flow. DocuDesk normalises the document, OpenAnonymiser finds the PII, OpenRegister stores the redacted version, OpenCatalogi publishes.

Sources
  • Word and PDF
  • Form submissions
  • OpenRegister records
  • Email and chat
01
Normalise
DocuDesk
Extracts plain text from Word, PDF, scans, and email. Hands it to OpenAnonymiser.
02
Detect
OpenAnonymiser
Finds PERSON, LOCATION, IBAN, BSN, ZAAKNUMMER, KENTEKEN, EMAIL. Returns spans plus confidence.
03
Redact
DocuDesk
Applies the spans. Stores both the redacted and the original alongside the audit log.
04
Publish
OpenCatalogi
The redacted version goes public. The original stays inside your Nextcloud, locked behind RBAC.
Consumers
  • Woo portal
  • Residents
  • Statistics and BI
Live in pilot

Public bodies and SMBs testing the pipeline.

OpenAnonymiser is in beta and a handful of organisations are running it in production today. Recall on Dutch PII is at parity with the upstream Presidio recognisers; the Conduction NER models lift recall on case numbers and licence plates that Presidio misses out of the box.

Two variants

Light or GPU. Pick by document volume.

The Light build runs on CPU only and is the right pick when document volume is under ~500 pages per day. The GPU build adds a transformer pipeline that lifts recall on longer documents but needs a GPU on the host.

OpenAnonymiser Light

OpenAnonymiser GPU