Government Algorithm Transparency

Short answer: Public authorities deploying high-risk AI systems are required under the EU AI Act (2024/1689) to register those systems in an EU database before putting them into service, inform individuals when a high-risk AI system is applied to them, and retain automatically generated logs for at least six months. These obligations apply from 2 August 2026. The Netherlands leads with a voluntary Algorithm Register currently listing 1,486 systems.

What the AI Act requires of public authorities

Regulation (EU) 2024/1689 — the EU AI Act — entered into force on 1 August 2024. Transparency and registration obligations for high-risk AI systems apply from 2 August 2026 (Article 113).

The regulation distinguishes two actors: providers (those developing or placing a system on the market) and deployers (those putting a system into use, including public bodies). Three articles are directly relevant for the public sector:

Article 13 — transparency toward deployers. High-risk AI systems must be designed so that their operation is sufficiently transparent. Providers must supply complete documentation: intended purpose and performance metrics, accuracy statistics, known risks to fundamental rights, instructions for interpreting outputs, and maintenance schedules.

Article 26 — obligations of deployers. Public authorities deploying high-risk AI systems must: (1) use the system in accordance with its instructions for use; (2) retain automatically generated logs for at least six months; (3) inform natural persons that a high-risk AI system is being applied to them; (4) notify the provider immediately upon identifying risks.

Article 49 — registration in the EU database. Public authorities must register themselves and the intended use of each high-risk AI system listed in Annex III in the Commission-managed EU database before putting the system into service. Systems used in law enforcement, migration, and border control are stored in a secure non-public section accessible only to the Commission and national authorities.

Which government functions are classified as high-risk (Annex III)

Annex III to the AI Act identifies the following public-sector applications as high-risk: AI for allocating essential public services and benefits (category 5a), AI in law enforcement and criminal risk assessment (category 6), AI in migration and asylum assessments (category 7), and AI in the administration of justice and democratic processes (category 8). AI in education (category 3) and critical infrastructure management (category 2) also fall within scope.

The Dutch Algorithm Register

Alongside the EU regulation, the Netherlands launched the Algorithm Register (algoritmes.overheid.nl) on 21 December 2022. This public register currently contains 1,486 algorithms from more than 500 government organisations. Registration remains voluntary; the Dutch government has announced that registration will become a legal requirement in the future for impactful algorithms, including high-risk AI systems. The Dutch Data Protection Authority (Autoriteit Persoonsgegevens) acts as the coordinating supervisory authority for algorithms and AI that pose risks to fundamental rights.