This week on Office Hours (https://officehou.rs), Dan Zitting and Kevin Legere discuss thoughts on risk management in scenarios that are "too good to be true". Kevin uses the Volkswagen emissions scandal to show how publicly available data and a machine learning technique illustrated the risk that emissions performance was just "too good to be true". He shows how simple unsupervised machine learning can identify such a situation when pointed at data that is being disclosed for compliance reasons anyway.