Drug-induced liver injury (DILI) is a major contributor to high attrition rates among candidate and market drugs and a key regulatory, industry, and global health concern. While acute and dose-dependent DILI, namely, intrinsic DILI, is predictable and often reproducible in preclinical models, the nature of idiosyncratic DILI (iDILI) limits its mechanistic understanding due to the complex disease pathogenesis, and recapitulation using in vitro and in vivo models is extremely challenging. However, hepatic inflammation is a key feature of iDILI primarily orchestrated by the innate and adaptive immune system. This review summarizes the in vitro co-culture models that exploit the role of the immune system to investigate iDILI. Particularly, this review focuses on advancements in human-based 3D multicellular models attempting to supplement in vivo models that often lack predictability and display interspecies variations. Exploiting the immune-mediated mechanisms of iDILI, the inclusion of non-parenchymal cells in these hepatoxicity models, namely, Kupffer cells, stellate cells, dendritic cells, and liver sinusoidal endothelial cells, introduces heterotypic cell–cell interactions and mimics the hepatic microenvironment. Additionally, drugs recalled from the market in the US between 1996–2010 that were studies in these various models highlight the necessity for further harmonization and comparison of model characteristics. Challenges regarding disease-related endpoints, mimicking 3D architecture with different cell–cell contact, cell source, and the underlying multi-cellular and multi-stage mechanisms are described. It is our belief that progressing our understanding of the underlying pathogenesis of iDILI will provide mechanistic clues and a method for drug safety screening to better predict liver injury in clinical trials and post-marketing.