On 11^th June 2026, the German Parliament (Bundestag) adopted the Act on the Implementation of the Artificial Intelligence Regulation. The implementing act designates the Federal Network Agency as the central market surveillance authority in Germany. This designation is based on Article 70(1) AI Act, which requires Member States to designate the market surveillance authorities responsible for implementing the AI Act. However, the supervisory structures provided for in the AI Act are complex, particularly with regard to the independence requirements applicable to the competent authorities.

Under Article 70(1) AI Act, second sentence AI Act, the competent authorities must exercise their powers independently, impartially and without bias. For certain areas of application of high-risk AI systems, the Regulation goes further: under Article 74(8) AI Act, “complete” independence of the competent supervisory authorities is required in particular in the field of biometrics (Annex III, No. 1) in connection with law enforcement, border management, justice and democracy. The same applies to the use of high-risk AI systems in the fields of law enforcement, migration, asylum and border control management, administration of justice and democratic processes (Annex III, Nos. 6, 7 and 8).

In addition, Article 74(3) AI Act provides that supervision of high-risk AI systems placed on the market independently or as safety components of regulated products pursuant to Annex I, Section A, falls within the responsibility of the competent sectoral market surveillance authorities.

These fragmented responsibilities, combined with differentiated independence requirements within the AI regulatory framework, raise fundamental questions concerning the practical implementation, institutional design and democratic legitimacy of AI supervision.

These questions are addressed by the interdisciplinary project group of the AI Supervision (AI-Sup) project, composed of scholars from law, economics, political science and philosophy. The project focuses on analysing the independence requirements of AI supervision and their compatibility with German and European theories of democratic legitimacy. To this end, comparable models of institutional independence in different legal systems are examined from legal, political science, economic and philosophical perspectives. In particular, the constitutionally anchored independence of the Bundesbank under Article 88 of the German Basic Law and the independence requirements applicable to data protection supervisory authorities under Article 52 GDPR are analysed with a view to identifying transferable parallels for the institutional design of AI supervision.

In addition, the project examines the interaction between public supervision and the private-law liability of market participants in order to outline a coherent architecture of responsibility within the scope of the AI Act.

The aim of the project is to develop viable institutional models that translate the EU-law independence requirements into supervisory structures that are compatible with constitutional and democratic principles.

The project group forms part of the Hessian-wide network Centre Responsible Digitality (ZEVEDI) and is led by Prof. Dr. Katja Langenbucher (Goethe University Frankfurt) as spokesperson. The deputy spokesperson was Prof. Dr. Florian Möslein, LL.M. (Philipps University Marburg) until April 2026 and has been Prof. Dr. Johannes Buchheim, LL.M. (Philipps University Marburg) since May 2026. Further members include Prof. Dr. Nathalie Behnke (TU Darmstadt), Prof. Dr. Petra Gehring (TU Darmstadt), Prof. Dr. Gerrit Hornung, LL.M. (University of Kassel) and Prof. Dr. Emanuel Mönch (Frankfurt School of Finance & Management).

Further information is available on the AI-Sup-website.

Funding: 
Centre Responsible Digitality (ZEVEDI)

Duration: 
January 2026 - June 2027

Project Leader: 
Prof. Dr. Gerrit Hornung, LL.M.

Staff: 
Argyro Tsapakidou