PrivaSearch

Law firms, notaries and bailiffs

AI for law firms and notaries, within the GDPR. With a duty of confidentiality that stands.

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client files with Claude, ChatGPT or Mistral Vibe, after names, BSNs (Dutch citizen service numbers) and addresses have been replaced by pseudonyms and a colleague has checked that. For a law firm, notary's office or bailiff's office, that means: querying litigation files and deeds with source references, without the AI knowing who your client is.

In short

A litigation file contains the data of your client, the opposing party and witnesses. None of them agreed to a chatbot.

PrivaSearch replaces names and numbers with pseudonyms, and a colleague releases every document.

An answer comes with the passage and the page, so you can find it again in the client file.

What you use AI for

The client files that matter.

A firm searches through stacks of documents in which everything has to be exactly right: who, when, which article. An assistant helps there, if it does not need to know the names.

  1. Litigation files

    Finding what a witness stated, which deadline was agreed or which position the opposing party took, with the page.

  2. Contracts and general terms and conditions

    Finding notice periods, liability clauses and penalty clauses in a folder full of agreements.

  3. Deeds and notarial files

    Searching deeds of transfer, articles of association and wills for a provision, without the AI knowing who they are about.

  4. Bailiff files

    Finding service of documents, attachments and payment arrangements per case, with the debtor's data replaced.

Which rules apply

What the law asks of your client files.

In a legal profession, confidentiality weighs heaviest. It requires that you know where every document goes, even when an assistant reads it.

The question we hear most often

Working GDPR-proof with AI at a law firm: what does that involve?

No tool does that for a firm, PrivaSearch included. The GDPR and the AI Act place the obligations with you as the controller. What your firm arranges itself, and what PrivaSearch does to help:

What your firm arranges itself

  • A lawful basis (Articles 6 and 9 GDPR)A lawful basis for every processing operation, and for special category data an exception from Article 9(2) as well.
  • A data processing agreement (Article 28 GDPR)With every party that processes data for you: with PrivaSearch, and with the provider of the AI assistant you connect.
  • A DPIA where one is required (Article 35 GDPR)Often with special category data or processing on a large scale.
  • An AI policyWhich tools are allowed, with which data, and who decides on that.
  • AI literacy (Article 4 AI Act)Staff who work with AI know what they are doing and what the risks are.
  • Professional secrecyThe duty of confidentiality of your profession remains yours. Per project, appoint only reviewers who may already view the client file.

What PrivaSearch does to help

  • Pseudonymising before the AINames, BSNs (Dutch citizen service numbers) and addresses are replaced before anything goes to your AI assistant.
  • A mandatory privacy reviewA colleague checks and approves every document. Only then is it searchable.
  • Permissions per projectOnly the members of a project see the documents of that project.
  • The activity logWho viewed, searched, released or revealed what, without document content in the log.
  • Storage in the EUOriginals are stored in data centres in the EU, in a database of its own per organisation.
  • An AI policy and a DPIA to start withStarter documents based on your own set-up, to have reviewed.

PrivaSearch is designed for the GDPR. The assessment of your situation remains yours and that of your data protection officer or lawyer.

How it works

From client file to answer, in four steps.

The AI only receives text that a colleague has released, with pseudonyms instead of names. See how it works.

  1. Add client files. Upload documents or connect a SharePoint library; SharePoint is only read.
  2. Detect and pseudonymise. A language model within PrivaSearch and Dutch patterns (BSN with the eleven test, IBAN, Dutch number plate, KvK number) find personal data and replace it with pseudonyms.
  3. A colleague reviews. Original and pseudonym are side by side. What the model missed, the reviewer adds; then they approve the document.
  4. Search with your assistant. Claude, ChatGPT Business or Mistral Vibe searches the released passages through a secure MCP connection, for example: “Which notice period is in the lease between [PERSOON_A4F2] and [ORGANISATIE_E349]?”

The difference

A client file in a chatbot, or through PrivaSearch.

Topic WithoutUploading a client file to a chatbot WithPrivaSearch between your client file and your assistant
What the AI sees The entire client file, with names, BSNs and addresses. Released passages with pseudonyms.
Check beforehand None. A mandatory review by a colleague.
Who may see it Whoever uploads it, outside any arrangement for the client file. Only the members of the project.
What is recorded Nothing. Every view, search and release in the activity log.
Where the originals are With the provider of the chatbot. In data centres in the EU.

What you type into the assistant yourself goes straight to it, outside PrivaSearch. So ask about the role or the pseudonym, not about the person.

Frequently asked questions

What you probably want to know.

Working GDPR-proof with AI at a law firm: what does that involve?

No tool does that for you. Your firm arranges a lawful basis, a data processing agreement, a DPIA where one is required, an AI policy, AI literacy and professional secrecy itself. PrivaSearch does the technology: pseudonymising before the AI, a mandatory review, permissions per project, the activity log, storage in the EU and a starter document for your AI policy and your DPIA.

May an advocate upload a litigation file to ChatGPT or Claude?

The entire client file then goes to the provider, with the data of client, opposing party and witnesses. That sits poorly with the duty of confidentiality and with data minimisation. With PrivaSearch, the assistant only searches released passages with pseudonyms.

Can the AI reveal a pseudonym as my client?

No. The link between pseudonym and person is kept in a separate vault with a key per organisation, and there is no function for the assistant there. Revealing is only possible in PrivaSearch itself, by a project member with write access, and every reveal goes into the activity log. Every project member can open the original in PrivaSearch; that is logged too.

Do I get source references with an answer?

Yes. A passage comes back with a short reference to the document and the page, so you can check the answer in the client file itself.

Does this also apply to a notary or a bailiff?

Yes. Every profession with a duty of confidentiality has the same problem with a client file in a chatbot. The permissions per project make sure that only those who belong to the case see the documents.

What does a colleague who is not on the case see?

Nothing of that project. Only the members of a project see the documents and the passages, and a connected assistant only searches projects the user is a member of.

Updated on

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 →