In this talk I will discuss the tenability of safety reliabilism, a version of reliabilism which states that a true belief is knowledge if and only if the belief is safe and reliably formed. First I will discuss which notion of reliability to adopt. I base the discussion around two versions of reliabilism: dispositional reliabilism and modalized reliabilism. I argue that we have strong reasons for preferring a dispositional account of reliability, in part because it allows us to remain reliabilists about justification. Reliabilism nonetheless fails to handle counterexamples involving veritic luck, which I take to be a sign that an anti-luck condition is required for knowledge (following, among others, Duncan Pritchard). Two of the most well-known anti-luck conditions are safety and sensitivity. A large portion of the talk will be dedicated to defending safety, which I will do by attempting to solve some of its well-known problems (as discovered  by Becker, Comesaña, Greco, and Roush). I will also argue that safety is preferable over sensitivity as a necessary anti-luck condition for knowledge by presenting a counterexample to Becker’s modalized sensitivity reliabilism. In sum, I hope to show that safety reliabilism is a plausible theory of knowledge capable of avoiding many (if not all) of the typical counterexamples we find in the literature to date.