Duration: August 2025 until June 2026
AI regulatory sandboxes are among the innovative procedural, administrative and supervisory tools designed to implement and further develop, on the basis of practical experience, the emerging AI governance and AI regulation in Europe. In a regulatory sandbox, selected companies can receive official support whilst developing an AI system. These procedures are intended to enable AI – as a ‘moving target’ – to be confronted with a regulatory framework that is itself adaptable, built upon regulatory and technical experience, and sufficiently flexible. Thus, EU legislation – specifically the AI Regulation, which comes into force on 1 January 2026 – grants companies opportunities for experimentation within the framework of these real-world laboratories: both in practical and technical terms and with regard to standards that may be critically affected (see Art. 57 et seq. AI Act). In addition to testing specific AI applications within a legally privileged, closely supervised framework under official guidance, the real-world labs are primarily intended to serve the generation and further development of AI-related risk knowledge and regulatory solutions.
The ZEVEDI ad hoc project carried out studies accompanying the process of the Federal Network Agency’s ‘AI Real-World Lab’ pilot project, in collaboration with the Federal Data Protection Commissioner and the State of Hesse, which used selected use cases to test a procedure for developing a concept for a practicable procedural instrument for Germany. ZEVEDI has documented the findings in a final report (published in German and English). For the scientific discussion, follow-up questions arise, including regarding the compatibility of innovation promotion and regulatory learning. The participating researchers have reported on this in a European Commission project and published a specialist article – ‘What colour are the authorities’ glasses? Incentive, target structure and conditions for success of real-world laboratories under the AI Regulation” (RDi 2026, 228).
Johannes Buchheim, Petra Gehring, Laura Grosser & Markus Schrenk
What colour are the authorities’ glasses?
Incentives, target structure and conditions for success of real-world labs under the AI Regulation
With its provisions on AI real-world labs, the AI Regulation is breaking new ground in terms of administrative procedure law and the culture of public authorities. The heterogeneous requirements for AI real-world laboratories raise many questions that authorities and AI providers must address. Among other things, it remains unclear what consequences arise from the fact that market surveillance authorities and notified bodies must ‘give appropriate positive consideration’ to the final report on tests conducted in the AI real-world laboratory. This is not the only reason why it remains an open question whether the specific details of AI real-world laboratories will be sufficiently attractive to AI providers to get the adaptive regulatory system of the AI Regulation, which they are intended to drive, off the ground.
Published in: Recht Digital (RDi) 2026, 228, pp. 228–234
Link to the article (in German)
“AI Regulatory
Sandbox Pilot Project”
Final report of the Bundesnetzagentur, the Federal Commissioner for
Data Protection and Freedom of Information and the Hessian Ministry for Digitalisation and Innovation | 29 April 2026
In a forward-looking pilot project, the Federal Network Agency (BNetzA), the Federal Commissioner for Data Protection and Freedom of Information (BfDI) and the Hessian Ministry for Digitalisation and Innovation (HMD) tested the implementation conditions for “AI regulatory sandboxes”. The European AI Regulation mandates the establishment of at least one AI regulatory sandbox in all Member States from August 2026. These are intended to provide for regulatory sandboxing (i.e. development and testing under regulatory supervision in a controlled environment to examine novel, complex legal issues with official oversight) and to facilitate the development by providers of innovative AI products that comply with the requirements of the AI Regulation and other relevant European legislation.
Based on two use cases, the pilot project was able to identify companies’ expectations of an AI real-world lab process, narrow down the scope of services offered by AI real-world labs, and clarify organisational issues relating to the establishment of such novel testing facilities.
‘AI Real-World Laboratory’ of the Federal Network Agency
Prof. Dr. Johannes Buchheim | Prof. Dr. Petra Gehring | Prof. Dr. Florian Möslein
Talk | 4 February 2026
At the exploratory workshop Fostering Academia–Public authorities co-operation for value creation: Mapping best practices and providing guidance for more effective knowledge uptake for policy making, organised in Brussels by the consultancy 4FRONT together with the research institute United Nations University – Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT) on behalf of the European Commission’s Directorate-General for Research and Innovation (DG RTD), the three project leads Prof. Dr. Johannes Buchheim, Prof. Dr. Petra Gehring, and Prof. Dr. Florian Möslein presented the Federal Network Agency’s pilot project “AI Real-World Laboratory.”
Project responsibility
Prof. Dr. Johannes Buchheim, LL. M. (Yale), Philipps-Universität Marburg | more information
Prof. Dr. Petra Gehring, Technische Universität Darmstadt | more information
Prof. Dr. Florian Möslein, Dipl.-Kfm., LL.M. (London), Philipps-Universität Marburg | more information
Research assistants
Laura Grosser, Technische Universität Darmstadt
Markus Schrenk, Philipps-Universität Marburg
