← Policy tracker Research · Policy

Netherlands publishes draft AI Implementation Act establishing decentralised national AI supervision framework

governanceenforcementregulationnational-authorityai-act

On 20 April 2026, the Dutch government published a draft AI Implementation Act (Wet Uitvoering AI-Verordening) for public consultation, closing 1 June 2026. The act formally establishes the national governance structure for AI Act compliance in the Netherlands, adopting a decentralised supervisory model: eight sectoral market surveillance authorities will oversee AI Act compliance within their respective domains, with the Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) and the State Inspectorate for Digital Infrastructure (RDI) acting as coordinating authorities. In total, ten bodies will share supervisory responsibility.

Each market surveillance authority will be able to impose administrative fines up to the maximum amounts set out in Article 99 of the EU AI Act, issue corrective orders, suspend AI systems, mandate market withdrawal, and issue public warnings. A legally mandated coordination protocol between all authorities is required to ensure consistent interpretation of key AI Act concepts across sectors.

The draft also assigns the AP the role of “notification coordinator” for high-risk AI systems and enables formal cooperation protocols with the EU AI Office.

Who it affects: All operators, deployers, and importers of AI systems in the Netherlands across all regulated sectors (health, financial services, transport, labour, media, etc.). Any organisation subject to existing sector-specific regulation will also face AI Act obligations through that sector’s designated authority.

Relationship to EU AI Act: The EU AI Act is directly applicable law in the Netherlands; this act creates the national machinery for supervision and enforcement. High-risk AI system requirements apply from 2 August 2026.

What is notably missing: The draft is still in consultation; it is not yet enacted. The decentralised model may create interpretive inconsistency across sectors despite the coordination protocol. The act does not introduce any AI literacy or training obligations beyond those already in the EU AI Act itself.