This webinar explores how trustworthy artificial intelligence in soil health monitoring is built—starting from modeling choices rather than ethical checklists. Focusing on AI-driven soil property mapping, the session shows how issues such as bias, fairness, transparency, and explainability emerge directly from data quality, model design, validation strategies, and uncertainty handling. The webinar then connects these modeling foundations to their ethical implications, including privacy and location sensitivity, social and environmental well-being, accountability, human agency, and inclusiveness. Drawing on practical experiences from AI4SoilHealth, it demonstrates how uncertainty-aware, explainable, and auditable models are essential for responsible use in policy and land management contexts. The session is aimed at researchers, modelers, and practitioners working at the interface of environmental modeling, AI, and decision support who want to design models that are not only accurate, but also trustworthy and policy-ready.