The Looming Closure of China's AI Frontier (www.aei.org)

China is on the verge of recalibrating its AI strategy amid rising concerns regarding the misuse of increasingly powerful AI models, particularly as tools for cyberattacks and terrorism. The announcement from Premier Li Qiang in 2024 of the “AI Plus” initiative aimed to democratize access to AI in support of economic growth, but recent developments highlight a troubling tension between fostering innovation and ensuring national security. Reports indicate that advanced AI models, such as Z.ai's GLM-5.3, have crossed critical thresholds that could facilitate damaging cyber activities, prompting Chinese officials to advocate for stricter governance. In response, Chinese AI labs are expected to impose restrictions on their more disruptive models while maintaining a focus on local, open-source applications. The newly introduced AI Safety Governance Framework underscores the risks associated with open models, particularly their potential for exploitation. As labs begin to limit access to their most potent technologies—attaching new, more restrictive licenses to their open-weight releases—there is uncertainty about how strictly these measures will be implemented. This shift suggests a bifurcation in strategy, with less oversight at the lower end of the AI value chain but increasing control over systems deemed too dangerous, marking a potential "soft" closure of China's AI frontier.