The increasing spread of generative AI tools poses considerable challenges for examination formats at universities. In particular, the question of individual performance is coming into focus, as AI is able to substitute many traditional examination services. This article takes a differentiated look at where AI plays a role in examination contexts, the resulting risks and which adjustments to examination formats are necessary and possible - both in the short and long term. Didactic as well as legal and organisational aspects are taken into account. The aim is to provide teachers with orientation and options for action in the area of conflict between examination law, didactic quality and technological dynamics.