Ein KI-Transparenzregister für die öffentliche Verwaltung

Abstract

Artificial intelligence (AI) holds the potential to become the cornerstone of digital transformation in public administration. While the use of AI systems promises more efficient operations through automation, it also introduces significant risks, particularly concerning data protection and the equal treatment of individuals. One of the most pressing challenges is the existing lack of transparency – both for citizens and for the administration itself. Outside the responsible government agencies, it is often unclear where and for what purposes AI systems are used. However, awareness of the systems in use and their functioning is critical to fostering public acceptance, ensuring effective oversight, protecting fundamental rights, and facilitating cross-agency knowledge sharing.

To enhance transparency regarding state usage of AI, the introduction of a public register is advisable – similar to the system already in place in the Netherlands. In the Dutch “Algoritmeregister”, government agencies are required to record the AI systems they use. The register provides key details such as the system’s intended purpose, its application domain, and the potential impacts on citizens and businesses. In early 2025, the German Federal Ministry of the Interior and Community took the first tentative steps toward this direction with the “Marktplatz der KI-Möglichkeiten”.

However, the establishment of a national AI transparency register for public administration raises significant legal considerations. The EU’s AI Act establishes its own framework for AI transparency, which takes precedence over national laws. According to the AI Act, public authorities must disclose the use of specific high-risk AI systems in a central database (Article 71 AI Act) and, notably, provide a summary of the fundamental rights impact assessment (Article 27 AI Act). If the AI Act comprehensively regulates AI transparency within public administration, the creation of a national register could conflict with EU law. Even if EU law does not completely preclude the creation of such a register, the approach to its implementation must be in line with German constitutional law. Specifically, the principle of the “Vorbehalt des Gesetzes” (reservation of the law) and Germany’s federal system of competences must be considered. This article explores these issues and outlines how the concept of a national AI transparency register can be legally implemented within the European multi-level system.