How aiNotaris measures up on the KNB AI-Weegschaal

What is the KNB AI-Weegschaal?

With the AI-Weegschaal, the Royal Dutch Association of Civil-law Notaries has created a practical assessment framework that helps notarial offices use AI safely. At its core is a weighing per application: what does using AI deliver and which risks stand against that. The annex to the framework couples that weighing to the statutory core values of the profession, including the confidentiality of article 22 Wna, and to principles for responsible use, such as outcomes that people can verify and oversight before, during and after.

The Weegschaal is deliberately neither a certificate nor a list of prohibitions. It is a framework that asks every office to choose consciously per application and to be able to explain that choice: to clients, to supervisors and to its own team. That aligns with what the law already requires, from the duty of confidentiality of article 22 Wna to the AI literacy duty of article 4 of the EU AI Act.

For a vendor, that implies a clear assignment: make the weighing easy. An office must be able to see, per theme, how the technology works and what evidence exists for it. Below, we therefore place aiNotaris on the Weegschaal along the three themes that weigh heaviest for the notarial profession: confidentiality, verifiability and human oversight.

Theme 1: confidentiality and professional secrecy

The weightiest theme for the notarial profession. With aiNotaris, privacy-sensitive data does not leave your own environment. The Data Shield anonymises personal data before processing: names, addresses and other identity markers are replaced before a language model sees anything, and the translation table stays with the office.

The environment runs on private Microsoft Azure infrastructure within the EU, not public cloud, and client data is never used to train AI models. That makes the confidentiality weighing an architectural choice rather than a promise: what the model does not know, the model cannot leak. How that construction works in detail is set out in our article on how we protect your files.

Theme 2: verifiability and traceability

An outcome that cannot be checked does not belong in a case file. That is why aiExpert works with structured source citations: legal research comes with references a case handler can look up, instead of a convincing-sounding answer without provenance.

In addition, every action is recorded in the audit trail. Who requested something, what the software proposed and who reviewed it: it can be reconstructed afterwards. For the Weegschaal that means a verifiable chain from input to outcome, and for the office it means a question from a client or supervisor can be answered.

Theme 3: human oversight

Human-in-the-loop is anchored in aiNotaris by design, not as a setting that can be switched off. The software proposes, the notary reviews and signs. That applies to drafts, to file checks and to every step that moves towards a deed.

In the workflow orchestrator, the office itself also records which steps may run automatically and where a person decides. Human oversight is thereby not a paper principle but a configured working agreement, different per process and per office where that is appropriate.

The evidence behind the weighing

A weighing is as strong as the evidence beneath it. aiNotaris is ISO/IEC 27001:2022 certified; the certificate is independently verifiable via IAF CertSearch. SOC 2 Type II and ISO/IEC 42001 are in progress. For a vendor assessment we supply the documentation on request, including a security description and a data processing agreement.

• ISO/IEC 27001:2022 certificate, verifiable in the IAF register.
• Description of the Data Shield and the anonymisation chain.
• Audit trail and roles and permissions, built into every layer.
• Data processing agreement and security documentation on request.
• SOC 2 Type II and ISO/IEC 42001: in progress.

The assessment remains your office's

To be clear: the KNB has not approved, certified or recommended aiNotaris, and it does not do so for any vendor; that is not how the Weegschaal works. The framework asks every office for its own, recorded assessment per application. What a vendor can do is make that assessment light: a clear answer per theme, with evidence you can verify.

So start with your own policy. Our free model AI protocol translates the AI Act, the GDPR and the Wna into workable office agreements and gives the Weegschaal a fixed place in your way of working. If you want to hold the themes against a live environment, we simply walk through them one by one in a demo.