Netherlands
EU-level rules also apply — The EU AI Act applies directly, with Annex III point 8(a) covering AI used by or for judicial authorities. The Rechtspraak states it follows the regulation 'als vanzelfsprekend' and excludes high-risk uses until it can meet the requirements. Dutch designation of national competent authorities (AP, RDI) was not verified against an official instrument in this pass.
Regulation Tracker · Country · Europe
Netherlands
Substantialcivil lawupdated 2026-09-04verified 2026-09-04
EU-level rules also apply — see the European Union record →
As of 4 September 2026 the Netherlands has the most client-facing bar guidance in this group: the NOvA's Aanbevelingen AI in de advocatuur, of December 2025, is the only European bar text telling lawyers to ask the client's permission before using AI on a file. A first disciplinary reprimand followed in July 2026.
As of 2026-09-04, the Netherlands has 2 binding rules, official guidance in 11 categories, 0 proposals and 1 position set only by case law across the 20 categories in the map. Binding rules cover horizontal AI statute, mandatory AI training. In 5 categories we found nothing; in 1 the position is not yet determined.
How to read this page
- Binding rule Binding rule means a rule in force from a court, statute, regulator or equivalent authority.
- Official guidance Official guidance means an official non-binding position from a judiciary, regulator, bar or public body.
- Proposed / consultation Proposed / consultation means a public bill, draft rule or consultation that is not yet in force.
- Set by case law Set by case law means no rule was found, but courts have set the position in decisions.
- Nothing found Nothing found means the category was checked and no rule, guidance, proposal or case-law position was found.
- Not yet determined Not yet determined means research did not establish a reliable position yet.
The rules, category by category
Article 4 of the EU AI Act binds law firms and courts as deployers from 2 Feb 2025; the NOvA recommendations state that AI use creates obligations under the AI-verordening and tell lawyers to keep knowledge current 'via permanente educatie' and to raise it in structured feedback for quality review. There is no dedicated mandatory AI PE point requirement.
NOvA recommendation Deskundigheid 3: 'Verifieer altijd de output — controleer citaten, jurisprudentie en feiten altijd handmatig vóór gebruik' and use only tools that cite sources. Enforced in Raad van Discipline 's-Hertogenbosch, 27 Jul 2026 (ECLI:NL:TADRSHE:2026:93) as an aspect of deskundigheid and integriteit under art. 10a Advocatenwet.
The Raad voor de rechtspraak's published AI strategy (from early 2025) sets a ten-point plan: focus on low-risk applications, 'learning-by-doing', an afwegingskader protecting judicial autonomy and art. 6 ECHR, independent oversight of that framework, and a public algorithm/AI register. Deployment of AI in high-risk processes stays excluded. RechtspraakGPT (released 10 Dec 2025) is limited to text tasks outside the primary process. The strategy followed criticism of Rechtbank Gelderland, 7 Jun 2024 (ECLI:NL:RBGEL:2024:3636), where a kantonrechter used ChatGPT to estimate the lifespan of solar panels and the average kWh price in assessing damages.
The Rechtspraak AI strategy point 9 states plainly: 'Rechterlijke oordeelsvorming (zoals een robotrechter) sluiten we echter uit.' It cites the AI Act and CCJE Opinion 26 on preventing improper AI influence on judicial reasoning. Not a court rule; there is no statutory ban.
NOvA 'Vertrouwelijkheid' recommendations: no confidential or client data in free or public AI models; privacy-by-design and documented trade-offs; know where data is stored and processed (country, sub-processors, security); keep input and output inside the firm environment; run a DPIA where personal data is processed; be aware of implicit data collection (location, open tabs).
NOvA Deskundigheid 1–2: basic knowledge of generative AI is 'essentieel voor iedere advocaat' — LLM principles, prompt engineering, bias mitigation, hallucinations, error detection, relevant regulation and cybersecurity — plus hands-on experience. Deskundigheid is a statutory core value (art. 10a Advocatenwet) and was the ground of the 2026 reprimand.
NOvA Onafhankelijkheid 3: 'Zorg voor toezicht op medewerkers en leveranciers, stel AI-beleid op en implementeer een controle cyclus.' Integriteit 2 requires a firm-wide AI policy. No binding supervision rule.
The strongest position in Western Europe: NOvA Vertrouwelijkheid 2 tells lawyers to 'Vraag toestemming aan de cliënt voor AI-gebruik in het dossier', and Integriteit 3 to inform clients of the firm's AI policy. Contrast Belgium (no duty to communicate) and Germany (no professional duty). Still a recommendation, not a gedragsregel.
NOvA 'Aanbevelingen AI in de advocatuur', published end of 2025 (reported 10 Dec 2025), prepared by the projectgroep Digitalisering & AI. Structured around the five statutory core values — deskundigheid, vertrouwelijkheid, onafhankelijkheid, integriteit, partijdigheid — with about seventeen concrete recommendations. The NOvA also points members to the CCBE guides.
The Rechtspraak AI strategy governs institutional deployment: low-risk uses (scheduling, pseudonymisation of judgments, inflow prediction and routing, B1-level summaries, chatbot public information, deadline checks, case-law search), a data-and-AI platform, compliance with the AI Act, transparency, and entry of every AI use in a publicly accessible algorithm/AI register. RechtspraakGPT is the in-house assistant.
The AP supervises algorithms and AI alongside the GDPR and publishes ongoing guidance (e.g. on preparing for the fundamental-rights impact assessment/FRIA, Aug 2026). No AP guidance specific to legal services located; the NOvA recommendations carry the DPIA requirement into the profession.
NOvA Vertrouwelijkheid 4 and Deskundigheid 4: check contract terms on data ownership, IP rights, liability, exit clauses and vendor lock-in; do not rely on FAQs or marketing promises but read the (privacy) terms; agree quality and safeguards with the supplier in advance; document which services process personal data.
Raad van Discipline 's-Hertogenbosch, 27 Jul 2026, ECLI:NL:TADRSHE:2026:93 — the first Dutch AI discipline case. A lawyer filed submissions in two rental disputes with fabricated or mismatched ECLI numbers (one pointing to a divorce ruling, another to a criminal case), then filed rectification acts without replacement authority. Breach of deskundigheid and integriteit (art. 10a Advocatenwet) and Gedragsregel 8. Sanction: berisping plus EUR 1,250 costs (EUR 750 NOvA, EUR 500 State); the confidentiality charge failed because no client data had been entered.
No landelijk procesreglement, court rule or practice direction requires disclosure or certification of AI use in filings as of 4 September 2026. The Rechtspraak AI strategy commits to developing 'visies, reglementen, afspraken' with the NOvA for handling AI in case material, but nothing is published.
No published court guidance for self-represented litigants using AI found as of 4 September 2026.
No evidence rule or guidance on AI-generated or deepfake material. The Rechtspraak AI strategy expressly flags AI-generated process and evidence documents and equality of arms as questions still to be worked out.
No rule, recommendation or reported case on billing for AI-assisted work found as of 4 September 2026.
Dutch law reserves only procesvertegenwoordiging in specified proceedings to advocaten (Advocatenwet); giving legal advice is not a reserved activity, so consumer AI legal products are not caught by a monopoly rule. No Dutch ruling or regulation on AI legal products found as of 4 September 2026.
Not researched — no Nederlands Arbitrage Instituut (NAI) AI rule or note identified before the search budget was exhausted.
Official documents recorded for Netherlands (1)
- Aanbevelingen AI in de advocatuur
NOvA · revised 2025-12-01 · verified 2026-09-04
Court incidents on the record in Netherlands
- Yichang District Court (Hubei)2026-07-21 Warning
Cite this page
SafeLegalAI Legal AI Regulation Tracker, "Netherlands" (updated 2026-09-04, verified 2026-09-04), https://safelegalai.com/regulation/country/nl (accessed 2026-09-04). Data: CC BY 4.0.
SafeLegalAI is a research publication by Cognesio LLP, not a law firm. Nothing here is legal advice, and no lawyer–client relationship arises from reading it. Rules change; always check the official document linked on each record and take advice on your own situation. Researched and drafted with AI assistance; verified against primary sources and edited by Cognesio LLP. The linked official documents are the record — our summaries are not the law and are not legal advice.
- NOvA — Aanbevelingen AI in de advocatuur (Dec 2025)
- Rechtspraak — Verantwoord en innovatief: AI voor een rechtvaardige Rechtspraak (AI strategy)
- Raad van Discipline 's-Hertogenbosch, 27 Jul 2026, ECLI:NL:TADRSHE:2026:93
- Rechtbank Gelderland, 7 Jun 2024, ECLI:NL:RBGEL:2024:3636 (kantonrechter used ChatGPT for factual estimates)
- Autoriteit Persoonsgegevens — Algoritmes & AI