This paper raises a problem for so-called safety-based conceptions of knowledge: It is argued that none of the versions of the safety condition that can be found in the literature succeeds in identifying a necessary condition on knowledge. Furthermore, reason is provided to believe that the argument generalises at least in the sense that there can be no version of the safety condition that does justice to the considerations motivating a safety condition whilst, at the same time, being requisite for knowledge.status: publishe