This essay motivates a revised version of the epistemic condition of safety and then employs the revision to (i) challenge traditional conceptions of apriority, (ii) refute ‘strong privileged access’, and (iii) resolve a well-known puzzle about externalism and self-knowledge.