Certifying artificial intelligence systems

The ethical and societal implications of artificial intelligence (AI) systems increasingly raise concerns. Trust is a crucial precondition for the acceptance and dissemination of AI. To create trust, rulemaking is necessary but not sufficient. It is at least equally important to promote compliance by designing suitable and effective mechanisms of enforcement. Among the different ex post and ex ante mechanisms with their various degrees of regulatory bite, certification (or ‘labelling’) provides for an intermediary approach to foster compliance. This chapter introduces certification mechanisms suitable for AI and discusses various options for their regulatory design.