Whenever the characteristics of a service grid environment are addressed, issues related to openness and dynamism pop out first. Such issues do affect the definition and handling of application exceptions, and traditional approaches to exception handling lack in proper mechanisms for capturing exception semantics and handling exceptions. In this paper, after analyzing the newly arisen problems of exception handling in a service grid environment, we focus on exceptions caused by runtime mismatches between user’s requests and underlying services, and propose a semantics-based approach to handling this kind of exceptions. The approach was first developed within the FLAME2008 project and some promising results have been achieved.