Digital twin models, similar to those used in aerospace engineering and manufacturing, could be utilized to inform global food policy and help predict and respond to systemic shocks in food production, processing, and consumption. Historically, major food emergencies have triggered advancements in food-system modelling, but fragmented efforts have limited progress. Advances in data generation, computing power, and AI offer the opportunity to create real-time, stress-tested models of food systems that can assess the impact of extreme weather, labour shortages, or export restrictions, as well as model the effects of multiple simultaneous stressors. However, implementing these models requires overcoming technical challenges, promoting data and model sharing, and ensuring strong governance and oversight.