Background There is a developing evidence base to support the use of risk assessment instruments in offenders with intellectual disability (ID). The aim of this study was to try to develop this literature with the inclusion of a control group of mentally disordered offenders without an ID, using the HCR‐20 and VRAG. Materials and Methods The VRAG and the HCR‐20 were completed for a group of offenders with an ID (n = 25) and a control group (n = 45), in four medium‐secure units across the UK. The outcome measure was physical aggression measured over 6 months. Results Both instruments consistently produced large effect sizes predicting any physical aggression and severe physical aggression. The structured clinical judgement based on the HCR‐20 was especially good. Conclusions The HCR‐20 and the VRAG have excellent predictive efficacy in offenders with an ID. A structured clinical judgement based on the HCR‐20 was especially predictive.