AI ethics are currently being done in a way that rely on methods of idealization—commonly used in ethics and philosophy—both in how individual issues are addressed but also in so far as the ethical landscaped is being distributed into subdisciplines of AI ethics (e.g., ethics of explainable AI and algorithmic fairness). This talk highlights how some of the features of this approach leads to challenges for ethical analysis, based on which he then sketches ways in which AI ethics can respond to these challenges.