When it comes to assessing breast cancer risk, most datasets that inform existing models come from white women, with consequences for using such models with non-white patient populations. Now, researchers show that an emergent AI-based predictor trained on people from diverse demographics can evaluate breast cancer risk accurately regardless of patients’ race or ethnicity. They tested its risk assessment accuracy by having it evaluate mammographs from more than 226,000 women with diverse demographic backgrounds. Breast cancer is among the most common cancers diagnosed in women. Risk models can predict patients’ likelihood to develop the disease, facilitating early detection, but most datasets behind existing models come from white cohorts. Consequently, they don’t predict breast cancer risk as well for populations of non-European origins. However, Shu Jiang and colleagues write: “Recent work on the use of mammogram risk scores (MRS), an AI-based mammogram image texture summary feature, developed in a racially diverse population, has shown considerable promise.” Now, they have further assessed this tool’s performance across two major North American cohorts. Data came from routine mammography screenings of more than 226,000 women, aged 40 to 74, and involved East Asian, South Asian, Indigenous, Non-Hispanic white, and Non-Hispanic Black demographics. Score values came from AI-based analyses of breast tissue’s texture and characteristics. The scores, which encompassed risk over a period of 5 years, were significantly associated with breast cancer risk and highly calibrated with outcomes for each racial and ethnic subgroup. “This work shows that the MRS is independent of self-reported race and ethnicity, which allows for universal clinical implementation of personalized risk assessment,” Jiang et al. conclude.