The study demonstrates the development and validation of an AI model, the AI-CTRCD model, which can predict cancer therapy-related cardiac dysfunction (CTRCD) from baseline 12-lead electrocardiograms (ECGs). CTRCD, a significant adverse effect of anthracycline chemotherapy, can be detected by the AI-EF model, which identifies reduced left ventricular ejection fraction (LVEF) from ECGs. The AI-CTRCD model was developed using transfer learning on the AI-EF model, leveraging its insights into ECG features reflecting left ventricular pathophysiology. In a cohort of 1011 patients treated with anthracyclines, 8.7% experienced CTRCD. High AI-CTRCD scores were significantly associated with an increased risk of CTRCD (hazard ratio [HR], 2.66; 95% confidence interval [CI], 1.73–4.10; log-rank p < 0.001). This association remained consistent after adjusting for known risk factors (adjusted HR, 2.57; 95% CI, 1.62–4.10; p < 0.001). The AI-CTRCD score improved prediction beyond known risk factors, with a time-dependent area under the receiver operating characteristic curve (AUC) of 0.78 for 2 years compared to 0.74 without the score (p = 0.005). The AI-CTRCD model robustly stratified CTRCD risk across various clinical subgroups, including cancer types, sex, baseline LVEF, and anthracycline dose. The model's performance was consistent across different institutions and subgroups, and it showed improved detection of CTRCD under limited echocardiogram capacity. The study highlights the clinical utility of baseline ECG and the AI-CTRCD model in identifying patients at high risk for CTRCD, potentially improving the management and outcomes of patients receiving cardiotoxic chemotherapy.The study demonstrates the development and validation of an AI model, the AI-CTRCD model, which can predict cancer therapy-related cardiac dysfunction (CTRCD) from baseline 12-lead electrocardiograms (ECGs). CTRCD, a significant adverse effect of anthracycline chemotherapy, can be detected by the AI-EF model, which identifies reduced left ventricular ejection fraction (LVEF) from ECGs. The AI-CTRCD model was developed using transfer learning on the AI-EF model, leveraging its insights into ECG features reflecting left ventricular pathophysiology. In a cohort of 1011 patients treated with anthracyclines, 8.7% experienced CTRCD. High AI-CTRCD scores were significantly associated with an increased risk of CTRCD (hazard ratio [HR], 2.66; 95% confidence interval [CI], 1.73–4.10; log-rank p < 0.001). This association remained consistent after adjusting for known risk factors (adjusted HR, 2.57; 95% CI, 1.62–4.10; p < 0.001). The AI-CTRCD score improved prediction beyond known risk factors, with a time-dependent area under the receiver operating characteristic curve (AUC) of 0.78 for 2 years compared to 0.74 without the score (p = 0.005). The AI-CTRCD model robustly stratified CTRCD risk across various clinical subgroups, including cancer types, sex, baseline LVEF, and anthracycline dose. The model's performance was consistent across different institutions and subgroups, and it showed improved detection of CTRCD under limited echocardiogram capacity. The study highlights the clinical utility of baseline ECG and the AI-CTRCD model in identifying patients at high risk for CTRCD, potentially improving the management and outcomes of patients receiving cardiotoxic chemotherapy.