In driving, Hazard Perception (HP) is defined as the ability to identify potential traffic hazards with sufficient time to avoid collision. Driving demands high visual, attentional, and memory engagement, involving executive functions such as sustained attention and inhibitory control. Even in autonomous driving, manual takeover may require several seconds without proper hazard perception. Executive functions and their neural correlates have traditionally been studied using artificial stimuli in static contexts, due to two main issues: eye movements generate artefacts in brain signals, and it is difficult to associate brain activity with dynamic, complex stimuli. In this work, we aim to apply new tools to study hazard perception in natural, dynamic driving environments. We will focus on the study of the association between hazard perception, eye movements and brain activity recorded with magnetoencephalography (MEG) while seeing UK driving test videos. We analysed fixation and saccade-related brain activity and started exploring associations between dynamic mental states identified using unsupervised learning methods and successful or failed hazard perception.This is a key step towards understanding brain responses to naturalistic driving situations.