Speculative execution attacks affect all modern processors and much work has been done to develop techniques for detection of associated vulnerabilities. Modern processors also operate on weak memory models which allow out-of-order execution of code. Despite this, there is little work on looking at the interplay between speculative execution and weak memory models. In this paper, we provide an information flow logic for detecting speculative execution vulnerabilities on weak memory models. The logic is general enough to be used with any modern processor, and designed to be extensible to allow detection of vulnerabilities to specific attacks. The logic has been proven sound with respect to an abstract model of speculative execution in Isabelle/HOL.