This study investigates the synergy of environmental risks and institutional quality in predicting non-performing loans (NPLs) focusing on 77 banks across 19 EU countries, operating under the Single Supervisory Mechanism (SSM). To this aim, we incorporate a comprehensive dataset of bank-specific, macroeconomic, environmental and institutional dimensions, spanning from 2005 to 2023. Utilizing Machine Learning (ML) methods, our key findings indicate that amongst the models employed, the Multilayer Perceptron (MLP) Neural Network yields the best performance. Additionally, the incorporation of both environmental and institutional quality significantly enhances the predictive accuracy concerning bank NPLs. Notably, institutional quality emerges as the most influential predictor, followed by environmental risks, underscoring its pivotal role in monitoring and mitigating external risks while also enhancing financial stability. These insights suggest that financial institutions should integrate environmental risks into their risk management frameworks while policymakers should implement robust institutional measures to safeguard financial stability.