PrivaSearch
Book a demo

How PrivaSearch works

Pseudonymise and check first. Only then AI.

PrivaSearch sits between your documents and the AI model. Personal data is detected within PrivaSearch itself and replaced by consistent pseudonyms; a person approves before anything becomes searchable.

PrivaSearch is software that lets your team search its own client files with Claude, ChatGPT or Mistral Vibe, after names, BSNs (Dutch citizen service numbers) and addresses have been replaced by pseudonyms and a colleague has checked that.

A minute and a half: what happens when a client file lands in a chat window, and how PrivaSearch answers the same question: detection, pseudonymisation, human review, search with source references and connecting through MCP.
  1. Collect Upload files or connect a SharePoint library. Read-only: nothing changes in your source systems.
  2. Detect A language model within PrivaSearch and Dutch patterns find names, addresses, BSNs (Dutch citizen service numbers), IBANs, Dutch number plates and dates of birth. No outside AI provider looks on at this step.
  3. Review A colleague sees the original and the pseudonymised text side by side, corrects what the model missed and explicitly approves.
  4. Search Questions such as "what is the notice period?" are answered with passages and source references, per project, only for those who have access.
  5. Connect Claude, ChatGPT Business or Mistral Vibe searches the same released passages through a secure MCP connection. Never the originals.

Limits of the model

What the AI does and doesn't see.

  • Does: released, pseudonymised passagesLimited in size, with a page reference, only from projects the user is a member of.
  • Doesn't: original files, file names or linksThe AI gets no access to your storage. In the standard configuration the AI layer has no route to the original.
  • Doesn't: who is behind a pseudonymThe link between pseudonym and person sits in a separate vault, with a key per organisation. Revealing is only possible in PrivaSearch, with a permission check and a log entry.
  • Search queries are filtered tooIf someone accidentally types a name into the PrivaSearch search field, it is replaced before the question reaches the model.
  • Does: what you type into a connected assistant yourselfIf you ask the question in your AI assistant instead of in PrivaSearch, that question goes straight to the provider of that assistant — bypassing our pseudonymisation. So don't put names in it, and only connect business accounts with a data processing agreement. How MCP and the GDPR fit together.

Access

Permissions follow your organisation.

  • Per organisation, per projectEvery organisation has its own database and its own address. Within the organisation, project membership determines who sees what; administrators get no automatic access.
  • SharePoint stays in chargeWith every request there is a live check whether the user may still see the document in SharePoint: when searching, when viewing, when downloading and through the MCP connection. Permissions revoked? Then it's gone here too.
  • Two-step verification and device confirmationMandatory for everyone. New devices are confirmed with a code by email.
  • Retention periods with teethAn end date per project. After that, access is blocked and copies, fragments and pseudonyms are cleaned up, unless a legal hold applies.

No training, no access

Everyone who searches knows what they're working with.

Employees take a short AI literacy training (in line with Art. 4 of the EU AI Act) with periodic refreshers. Without a valid training the workspace doesn't open and the MCP connection refuses too.

See in twenty minutes what happens to your documents.

We show it with fictitious client files from your sector: upload, detection, review, search and the connection with your AI assistant. Afterwards you'll know exactly what your team can use safely.

Book a demo →