Can a notary use ChatGPT?

So is it allowed or not?

The question comes up at virtually every notarial office: can a notary use ChatGPT? The answer is nuanced, but not vague. Yes, ChatGPT and comparable public AI tools can be used for general tasks that involve no confidential information. No, client or case file data never belongs in them, not even partially and not even as a one-off. Practice also shows that the use is already there: staff try it at home or at the office, with or without the notary's knowledge. That is exactly why the question deserves a clear answer rather than tacit tolerance.

The line is identifiability. As soon as a prompt contains names, addresses, citizen service numbers, amounts, family relationships or other case file details, the office is sharing confidential information with an external party over which it has no control. That goes straight to the duty of confidentiality in Article 22 of the Wet op het notarisambt (Wna), which makes no exception for convenient tools.

In our whitepaper IT-Landschap Notariaat 2026 we classify public chatbots such as ChatGPT at the lowest level of AI maturity: generic tools without notarial context and without a Data Shield. Our verdict there is strict and deliberate: using them for client files is indefensible under notarial disciplinary law.

Why does case file data never belong in a public chatbot?

Three legal frameworks converge the moment case file data enters a public AI tool.

• Article 22 Wna: the duty of confidentiality covers everything the notary learns in the exercise of the office. Entering it into an external chatbot means sharing it with a third party, without any notarial basis for doing so.
• The GDPR: a prompt containing personal data is a processing operation. With public tools it is often unclear where data is stored, whether it is processed outside the EU and whether prompts are used to train models. Without a processing agreement and a legal basis, that cannot be justified.
• The duty of care: the notary must stand behind the accuracy of their work. Public chatbots can produce convincing but incorrect answers, without source citations and without traceability.

Added to that, free and consumer variants of these tools may use prompts by default to improve their models. Whatever has been entered cannot be retrieved. For a profession whose very existence rests on confidentiality, that is an unacceptable risk.

What does the KNB AI-Weegschaal ask?

With the AI-Weegschaal, its AI balancing framework, the KNB has set out a practical assessment framework for the safe use of AI in the notarial profession. The core: weigh the benefits against the risks for each application, and safeguard confidentiality, verifiability of outputs and human oversight throughout.

For ChatGPT use, that means in concrete terms: define what the tool may and may not be used for, make sure staff understand what happens to a prompt, and record that outputs never end up in a case file or deed unchecked. A tool that offers no insight into its sources and workings quickly fails that weighing the moment anything confidential is at stake.

Document the weighing as well. An office that can show when and why an AI application was permitted or rejected stands stronger towards clients, supervisors and its own team.

What does Article 4 of the EU AI Act mean for your office?

Since 2 February 2025, Article 4 of the EU AI Act (Regulation (EU) 2024/1689) has applied: organisations that use AI systems must ensure a sufficient level of AI literacy among their staff. That duty also applies to a notarial office where staff use ChatGPT, even occasionally.

AI literacy means that staff understand what an AI system can do, where it fails and what risks its use carries. An office must be able to demonstrate that it has taken this seriously. A written AI policy with accompanying instruction is therefore no longer a luxury, but the logical way to meet a statutory duty. The Gedragscode Informatiebeveiliging Notariaat, the profession's code of conduct on information security, points in the same direction: offices must have their information security demonstrably in order, and unregulated chatbot use does not fit that picture.

Which tasks can go into ChatGPT, and which never?

In its article on when not to use generative AI, Gartner advises restraint wherever factual accuracy, explainability and confidentiality weigh heavily, and recommends deploying generative AI mainly where language skill and variation are wanted. Translated to notarial practice, that produces a clear division. The rule of thumb for the office: anything you would not publish on a public website does not belong in a public chatbot either.

Without client or case file data, a public chatbot can for instance help with:

• Rephrasing general texts, such as an explanatory paragraph for the office website.
• Brainstorming the outline of a presentation, newsletter or internal instruction.
• General questions unrelated to any matter, where the answer is always verified independently.

Never belonging in a public chatbot:

• Draft deeds, clauses or correspondence containing party details.
• Case file documents, not even summarised and not even partially anonymised.
• Wwft assessments, risk flags and anything touching the verification of a client's intent.
• Legal conclusions that the client or the office must be able to rely on without independent review.

What does a safe setup look like?

A ban alone does not work. Staff reach for ChatGPT because the work keeps piling up, not out of ill will. A workable policy therefore combines three elements: clear rules, a safe alternative and human oversight.

Start with an AI protocol that records which tools are permitted, for what purpose, and what must never go into a prompt. We have drawn up a free model AI protocol for exactly this, translating the AI Act, the GDPR and the Wna into workable office rules. You adapt it to your office and adopt it today.

Then offer a private alternative that can handle case files. At aiNotaris, privacy-sensitive data never leaves your own environment: the Data Shield anonymises personal data before processing, the environment runs 100% privately within the EU and aiNotaris is certified to ISO/IEC 27001:2022. How that works is set out on our confidentiality and security pages. And because aiNotaris works alongside your NSL/DMS, no one needs to copy and paste into a separate chatbot any more: the digital colleagues work inside the case file itself.

Finally, keep the human in charge. The software proposes, the notary reviews and signs. That human-in-the-loop principle is not only a premise of the KNB AI-Weegschaal, but also the only setup that fits a public office in which responsibility cannot be delegated.