This chapter argues that, at least in epistemology, it is better to be safe than sensitive. It looks to adjudicate between safety theories and sensitivity theories. Each theory tries to explain away its counterintuitive results, but there has been little consensus regarding the success of such efforts. Strong Restricted Sensitivity and Strong Safety conditions are extensionally equivalent within a restricted space of worlds. The chapter argues that considerations about the "point and purpose" of the concept of knowledge motivate a standard safety condition rather than a standard sensitivity condition. Specifically, considerations regarding the close relationship between knowing and being a good informant suggested that knower must be reliable throughout a space of close counterfactual situations. This chapter looks for a non-standard sensitivity condition that could accommodate this line of thought, making use of the notion of a restricted set of worlds.