Signal Military AI governance under strain: the US–China dialogue

Summary

Days after the Trump-Xi summit in Beijing on 14-15 May 2026, at which a bilateral channel on AI governance was announced, China's delegation to the 23rd IISS Shangri-La Dialogue in Singapore (29-31 May 2026) called for rules, potentially legally binding ones, to fill the regulatory vacuum around military AI. The Chinese delegation head, Major-General Meng Xiangqing, stressed that final authority over weapon systems must remain in human hands. However, China framed its call within the UN-centred multilateral framework it favours rather than through the new bilateral channel with Washington. China has not endorsed the human-control commitments of the Responsible AI in the Military Domain (REAIM) summits and declined to sign the 2024 REAIM Blueprint for Action reaffirming human control over nuclear-weapons decisions. Its November 2025 arms-control white paper instead framed human control around 'relevant weapons systems' generally, a broader and more permissive formulation than nuclear-specific language. Analysts note this rhetorical convergence coexists with an assessment that the agreement excluding AI from nuclear decisions is weakening in practice.