Project Details
Projekt
Potentials of the certification of AI systems in regulating Artificial Intelligence
Subject Area
Principles of Law and Jurisprudence
Public Law
Public Law
Term
since 2025
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 547583482
The project aims to clarify the legal basis for the certification of AI systems on the basis of interdisciplinary and comparative law work. The methods of artificial intelligence, in particular machine learning, harbour enormous opportunities for progress and prosperity. However, the use of AI systems developed through machine learning also entails risks that require appropriate risk management and a corresponding legal framework. A key element of risk management for AI systems is conformity testing and certification. Current legislative proposals, namely the European AI Act and the Brazilian draft AI Act, explicitly require the certification of AI systems. Of even greater interest are voluntary certifications that can offer legal or economic advantages. However, the legal significance and legal framework of such certifications are still largely unclear. Certifications can have very different legal significance, for example as a prerequisite for market access, but can also bring other legal advantages, for example to avoid liability. The requirements for the organization of certification procedures also vary greatly. The European AI Act, for example, differentiates between self-certification and certification by third parties. The certification of AI systems can be very costly in organizational and financial terms, depending on how it is structured. Limiting certification to what is necessary (such as the scope and intensity of testing) is a key prerequisite for the successful regulation of certification for AI. In view of the enormous diversity of certifications, the pivotal question of the project is what type and form of certification is appropriate for the purposes of the respective legal regulation that refers to certifications or attaches importance to them. The project analyses the legal significance of certification both with explicit reference and with "implicit" reference (relevance of certification), which is relevant in liability law, for example, on the basis of existing law and current legislative proposals (AI Act). Analyzing the relevant legal regulations will work out the general elements of certification. The results obtained on German and European law are compared with the results of the sister project on Brazilian law and general principles of legal significance are derived from them. In the technical part of the project, we will analyze current approaches to conformity testing and certification of AI systems and evaluate them with regard to their suitability for legal purposes. On the basis of these analyses, we will then develop proposals for an efficient certification of AI systems, which can make a significant contribution to risk limitation and legal certainty.
DFG Programme
Research Grants
International Connection
Brazil
