Many different instruments have been developed to assist in the assessment of risk for violence and other criminal behavior. However, there is limited evidence regarding how these instruments work in the ‘real world’. Even less is known about how these instruments might work for assessing risk in jail diversion populations, whether in research or practice. To address these knowledge gaps, the present study examined the characteristics of risk assessments completed by program staff (n = 10) on 96 mental health jail diversion clients (72 men and 24 women) using the Short-Term Assessment of Risk and Treatability (START). The findings provide preliminary support for the reliability and validity of START assessments completed in jail diversion programs, the first evidence of the transportability of START outside psychiatric settings, and further evidence regarding the reliability and validity of START assessments completed in the field. They additionally support the consideration of an eighth, general offending risk domain in START assessments. Copyright © 2012 John Wiley & Sons, Ltd.