Structured guidelines are fundamental for risk assessment, formulation and decision-making in medium secure forensic psychiatric services in the UK. The most commonly used guideline for violence risk assessment in the UK is the HCR-20, so it is important that the new version of the HCR-20 (HCR-20V3) is validated in UK samples. The aims of this study were to investigate if the HCR-20V3 has satisfactory interrater reliability and predictive validity for community violence. A prospective confidential inquiry design was used. The HCR-20V3 demonstrated very good inter-rater reliability and significantly predicted community violence at six and twelve months post-discharge, with ROC AUCs of.73 and.70 respectively. Implications for future research and practice are discussed.