China's shikong AI risk framework

China's concept of 'shikong' (loss of control) organizes AI risk around the identity of the actor losing control, not the severity of harm or the mechanism by which it occurs. Zilan Qian and the Oxford China Policy Lab identified three distinct loci in Chinese discourse: an operator losing control of an AI system, the state losing control, and humanity permanently losing control. Qian warns Western observers against over-indexing on keyword mentions without this contextual understanding, and against assuming rhetorical overlap signals genuine alignment on AI safety between China and the West. One specific gap the analysis identifies is that China lacks discourse about loss-of-control risks within internal lab deployments, which the analysis describes as one of the biggest dangers. Separately, Jasmine Li mapped China's AI safety research ecosystem, estimating fewer than 100 full-time frontier safety researchers in China and approximately $20 million in annual philanthropic funding, compared to over 1,000 researchers and more than $1 billion in the West.