Abstract: Food security is a strategic priority concerning national development and people’s well-being. Enhancing grain production resilience has become a core task for safeguarding national food security in the new era, while policy-based agricultural insurance serves as a key policy instrument for agricultural risk management. Based on panel data from 31 provinces in China covering the period from 2002 to 2024, this study employs the H-P filtering method to measure grain production resilience and applies a double machine learning model to systematically examine the impact, transmission mechanisms, and regional heterogeneity of policy-based agricultural insurance on grain production resilience. The results show that policy-based agricultural insurance significantly enhances grain production resilience, and this conclusion remains robust after a series of robustness tests. The main transmission mechanisms operate through promoting the optimization of cropping structure and encouraging investment in agricultural machinery. Moreover, the policy effects exhibit significant regional heterogeneity, with stronger effects observed in major grain-producing areas, grain production-sales balance regions, regions with high climate risks, and provinces with high dependence on foreign trade. Accordingly, this study proposes policy recommendations, including optimizing the premium subsidy mechanism, strengthening policy support for staple grain production, deepening policy empowerment, and implementing differentiated regional insurance policies. The findings provide references for improving the agricultural insurance system and strengthening the national food security framework.