The development of autonomous driving systems mandatorily requires solving two tasks: A promising implementation must be developed. Its safety must be demonstrated before it is placed on the market. The prerequisite for solving the second task is a profound understanding of the large area of product safety. In order to develop this, both a suitable nomenclature and a taxonomy are proposed for the first time. These form the basis for further considerations on the question of product validation, which show that the possibilities in this respect are very limited. Based on this interim result, the concepts of quantitative behavioral and product safety are derived and proposed, which demonstrably form the only possible foundation of any safety argumentation for autonomous vehicles. The former task, namely the design of promising implementation, is deliberately omitted. Complementary considerations, hopefully sparking new drive and at least regarded as helpful, are reserved for a separate book, which is in preparation.