Artificial Intelligence-Enhanced Network Modelling of ESG Risk in Global Supply Chains

Environmental, Social and Governance (ESG) risk is increasingly influenced by inter-firm relationships embedded in global supply chains. Challenging firm-level approaches that treat ESG exposure as independent across companies. This study examines whether firms’ structural positions within supply-chain networks are associated with ESG risk exposure and whether incorporating network information improves ESG risk prediction. The analysis draws on an international dataset integrating validated supplier-buyer relationships, shipment-level trade data. ESG incident records and sentiment derived from ESG-related news. Network-based econometric models and graph-oriented learning approaches are evaluated against conventional firm-level benchmarks. The results indicate that ESG risk clusters within connected groups of firms, with higher exposure observed among firms occupying central or intermediary positions in supply networks. In addition, ESG-related media sentiment exhibits predictive power for subsequent ESG incidents, supporting its role as an early warning signal. Overall, models that explicitly account for network structure deliver more accurate and better-calibrated predictions than standard econometric and machine-learning approaches. These findings highlight the value of a network-informed perspective for ESG risk assessment in complex international production systems.