The safe navigation of ships is of high societal concern. A promising approach for analysing waterway risks is using non-accident events as surrogate indicators of collision accidents. Such non-accident events are typically detected in data from the Automatic Identification System (AIS). Recognizing the significant interest in this approach, this article provides a review and analysis of methods based on the detection of non-accident events from AIS data, which aim to provide insight in maritime waterway risk. Considering also recent calls for increased focus on foundational issues in risk research and safety science, all related literature is reviewed in detail, and analyzed based on five questions: How are non-accident events defined, i.e. what terminology is used? What is the accident-theoretical basis of the method? How are non-accident events ranked? How is the method used? To what extent has the method been validated? Based on the results, it is concluded that focus is needed to build evidence of the validity of the models’ results, if these are to be effectively used for waterway risk analysis. As a prerequisite, focus is needed on how exactly non-accident events are defined, and what factors are involved in the relation between their occurrence and accident involvement.