This study provides a framework for evaluating and prioritizing artificial intelligence (AI) risks in military command and control (C2) cycle, supporting more informed decision-making and risk mitigation strategies. As AI increasingly transforms various aspects of our lives, it offers significant advantages in military operations by improving the speed and accuracy of decision-making. By exploiting a wide range of battlefield enablers, including sensor data, imagery analysis, electronic warfare inputs and secure communications, AI can process real-time operational information and propose meaningful decisions, thereby minimizing human-induced errors and enhancing the speed and effectiveness of dynamic military operations. However, this integration brings critical risks, including ethical concerns, security vulnerabilities, accountability and oversight. Therefore, AI implementation in C2 still must be carefully managed, ensuring continuous human oversight. While the literature on military AI applications is growing fast, there remains a significant gap in quantitative assessments of AI-related risks in military decision-making. Existing studies often emphasize technological advancements or doctrinal and ethical concerns, with little focus on a structured risk assessment. To address this, we apply the Spherical Fuzzy Z-Number Best-Worst Method, a hybrid multi-criteria decision-making approach, that combines the efficiency of the Best-Worst Method with spherical fuzzy sets to capture ambiguity and Z-numbers to reflect expert confidence levels. Our analysis identifies and compares seven key risks in AI-supported C2 decision-making, with “Security and threats” ranked the highest and “Output-related issues” considered the least critical.