Validated genomic classifiers are recommended for use when this may change the management of patients with prostate cancer (PCa). However, their potential for management de-escalation is unclear. Here, we retrospectively evaluated the 17-gene Genomic Prostate Score (GPS) in patients with clinically localised and locally advanced PCa (n = 409; median follow-up ≥6 yr) managed with active surveillance (AS), radical prostatectomy (RP), or radiotherapy (RT). To evaluate de-escalation potential against metastasis and European Association of Urology (EAU) high-risk biochemical recurrence (BCR), patients with low GPS scores (58% of the study population) were stratified into "concordant" and "discordant" low-risk groups based on agreement with standard National Comprehensive Cancer Network (NCCN) risk categories. The "discordant low-risk" phenotype (NCCN ≥2 but low GPS) proved highly prevalent, comprising 46% of NCCN≥2 AS, 57% of NCCN≥3 RP, 36% of NCCN≥4 RT, and 48% of NCCN5 locally advanced patients. Six-yr event-free survival estimates in these patients were 88% in AS, 100% in RP and RT, and 95% in the locally advanced groups, similar to the 100% 6-yr treatment-free survival seen in all "concordant low-risk" (low GPS and lower-risk NCCN) patients. Overall, identifying a prevalent group with favourable 6-yr outcomes despite high clinical risk supports the feasibility of future trials using genomic classifiers for targeted management de-escalation.