AI in the medical context is interesting in that it brings together two different ethical perspectives. In medicine the traditional framing is that of the ethical obligations of the doctor towards the patient, and the main aim is to provide the best treatment for the patient: to act in the patient’s best interest – to do what is good for them, to do what they want to have done to them, to act justly and to not do any harm. AI, as a technology, however, also raises all the issues that we typically ask in ethics of risk: Is it safe? Can it be misused? Are there unwanted outcomes that we might overlook? The main ethical considerations here are those of safety, robust functioning, but also external effects, unwanted side-effects, and other issues related to who actually benefits, who is made vulnerable and exposed to potential risks, and who will be in charge and make the decisions. From an ethics of risk perspective, the main concern is to avoid catastrophe – to avoid unforeseen risk of harm – to the environment, to human health and lives, ands o on. The question is: what kind of moral framework do we need to capture all relevant ethical concerns when we look at AI applications in the clinic? I shall argue that we need both ethics of risk and medical ethics – and that combining the four principles from biomedical ethics plus explainability will not suffice.