This study examines the validity of the COMPAS with offenders who have a serious mental illness. A widely used risk-needs assessment tool, the COMPAS was found to be a good predictor of re-arrest with this population, although it was more effective in distinguishing low-risk offenders from all others than in identifying those who pose a medium as opposed to a high risk of re-arrest. Overall, approximately two-thirds of study-participants were classified as low risk. At the same time, more than half of the sample scored in the high range on the important needs domains of substance abuse, criminal personality, and criminal thinking. The study discusses potential implications for using the COMPAS with mentally ill individuals.