I am pretty sure that deliberately tricking an automated system into causing bodily harm is already coved by existing law. Think of all the automated systems that have existed before ML.