Individuals acquitted as not guilty by reason of insanity (NGRI) are usually committed to psychiatric hospitals for treatment until they are considered suitable for conditional release back to the community. The clinical evaluations that inform conditional release decisions have rarely been studied but provide an ideal opportunity to examine the reliability and validity of complex evaluations in the field. For example, to what extent do forensic evaluators agree about an acquittee’s readiness for conditional release? And how accurate are their opinions? We reviewed 175 evaluation reports across 62 cases from Hawaii, which requires 3 separate evaluations from independent clinicians for each felony NGRI acquittee referred for conditional release evaluation. Evaluators agreed about an NGRI acquittee’s readiness for conditional release in only 53.2% of evaluations (κ = .35). Courts followed the majority evaluator opinion in 79.3% of all cases but ruled in an opposite direction from the majority evaluator opinion in more than a third of cases in which evaluators disagreed. Evaluators accurately differentiated those conditionally released acquittees who remained in the community from those who were rehospitalized in 62.4% of cases. Among the 43 insanity acquittees who were ultimately released, evaluator agreement was significantly associated with rehospitalization within 3 years. When the evaluators unanimously agreed that conditional release was appropriate, only 34.5% were rehospitalized. When the evaluators disagreed, 71.4% were rehospitalized. Overall, results reveal poor agreement among independent evaluators in routine practice but suggest that opinions may be more accurate when evaluators agree than when they disagree.