Generative AI (GenAI) has increased human acceptance towards randomized AI output, maybe trusting random responses to ethical dilemmas. We argue that randomization is relevant in AI decision-making for three main reasons: (1) From a machine learning (ML) perspective, randomized classifiers can also be optimal in the sense of minimizing error in unseen instances. (2) From a game-theory standpoint, randomization is the only approach to obtain a Nash Equlibira (optimal strategy) in some settings. (3) Randomization aligns with responsible AI (RAI) principles and emerging AI regulations, supporting fairness, while also offering transparency, thereby fostering accountability.