The increasing implementation of artificial intelligence into industrial and social settings requires legal regulation. Even after the publication of the first legal framework for AI by the EU Commission, industry and science are facing legal challenges. On September 14, Professor Dr. Georg Borges addressed the potential of testing and certification procedures for AI systems in his Webtalk under the heading "Safe AI systems through certification?". The slides of Prof. Borges presentation can be found here.
The EU Commission's regulatory proposal includes a risk classification of AI applications and sets out different requirements for companies depending on the classification of risk. In his presentation, Professor Borges outlined the scope and complexity of the corresponding tests and the parameters that need to be taken into account to determine a companys risk classification.
Thus defining requirements for possible tests and certification procedures for AI-based systems form the core task of the "ExamAI" project. The goal is to make the operation of AI systems comprehensible, measurable and therefore certifiable. Further participants of the panel discussion were Cornelia Kutterer, Microsoft, Dr. Johanna Wenckebach, Hugo Sinzheimer Institute for Labor Law (HSI) of the Hans Böckler Foundation and Irina Orssich, European Commission. The webtalk was moderated by Nikolas Becker, Gesellschaft für Informatik.
Further information on the project can be found here.