Despite everyone's best efforts, things will sometimes go wrong with AI systems, and the parties involved may find themselves in litigation or arbitration. When this happens, the parties will need to peer into the black box to reconstruct what the computer did and why it did it. But the usual electronic discovery tools and evidentiary principles for electronic records might not work for an AI system, especially a machine learning system that continuously learns based on new data. During this episode Eric Evans, Reginald Goeke and Alex Lakatos discuss some of these challenges.