Researchers have demonstrated that artificial intelligence analysis of mammograms reveals dynamic changes in breast cancer risk scores over time, offering a novel approach to predicting future disease before clinical symptoms emerge. Published in Radiology, the study led by Constance D. Lehman, M.D., Ph.D., professor at Harvard Medical School and CEO of Clairity Inc., establishes that image-based risk trajectories differ significantly between women who develop cancer and those who remain disease-free. The investigation analyzed serial mammograms from a cohort of 54,014 women undergoing screening between 2009 and 2019 across six diverse imaging sites. Utilizing a validated deep learning model trained on entire mammographic images rather than predetermined features like breast density, researchers generated continuous five-year risk scores for 158,807 exams. The final analysis compared outcomes for 817 women diagnosed with breast cancer within 365 days of their index exam against 53,197 cancer-free controls. Results indicated that AI-derived risk scores for cancer patients increased progressively over a six-year window preceding diagnosis. The median risk score rose from 2.1 in the early study period to 6.6 at the index exam, with the steepest increase occurring two years before diagnosis. In contrast, scores for cancer-free women remained stable, ranging between 1.8 and 2.2 throughout the observation period. These changes were detectable as early as six years prior to diagnosis, representing signals invisible to the human eye. The findings are particularly relevant for the majority of breast cancer cases, which are sporadic and lack significant family history or known genetic mutations. Traditional risk models often exhibit limited discrimination in such populations. By capturing longitudinal image-based changes, this dynamic biomarker approach provides risk assessment grounded in imaging data, potentially mitigating disparities associated with self-reported clinical information. Lehman emphasized that the evolving risk scores function as a new domain for preventive care, analogous to monitoring cholesterol or hypertension. The ability to identify predisposition through imaging alone opens opportunities for targeted, risk-reduction strategies. AI-integrated risk assessment is gaining clinical traction. The 2026 National Comprehensive Cancer Network guidelines now incorporate image-based risk scores, recommending additional breast MRI screening for women aged 35 and older with elevated five-year risk scores exceeding 1.7%. An FDA-approved AI model for five-year risk scoring is already in use at selected healthcare institutions across the United States, signaling a shift toward routine, dynamic risk stratification in breast cancer screening programs.