Suppose you accept a moral norm according to which it is wrong to harm a person. How should you then reason about a case in which someone will be harmed with a probability less than one? How exactly are we to draw the line between morally acceptable risks and risks that should not be accepted? The aim of this chapter is to suggest a novel answer to this question, which we believe can be rendered compatible with both consequentialist and non-consequentialist positions in normative ethics.