Military applications of artificial intelligence (AI) are becoming a central but poorly understood source of mistrust in US–China strategic competition. AI does not simply intensify rivalry by producing new capabilities; it also obscures intent, compresses decision-making and magnifies worst-case assumptions. Because many AI systems are dual-use, software based, commercially developed and technically opaque, each side struggles to distinguish defensive modernisation from preparations for coercion or pre-emption. The United States views military AI as essential to decision advantage and operational resilience, while fearing China’s civil–military fusion and ‘intelligentised’ warfare. China, in turn, sees US AI-enabled intelligence, surveillance and reconnaissance, uncrewed systems and export controls as threats to its security and technological development. These competing narratives reinforce instability. Modest risk-reduction measures, including regular military-to-military dialogue, shared benchmarking, cooperation on interpretability and guard rails for AI in nuclear-related decision systems, would diminish it.