The US State Department is preparing to investigate American companies' growing reliance on Chinese-made AI models, warning that systems like GLM 5.2 from Z.Ai and Kimi 2.7 from Moonshot are "designed to advance Beijing's narratives, censor dissent and reflect CCP ideology." Cost pressure is the primary driver: Coinbase CEO Brian Armstrong confirmed the exchange uses both Chinese models, while startup Lindy switched to Deep Seek to contain expenses. Nikkei Asia reports Airbnb and Uber are also racing to adopt Chinese alternatives.

The surge correlates with Anthropic suspending its Mythos 5 and Claude Fable 5 models at government request, leaving enterprises scrambling for replacements. Yet a sweeping ban faces steep obstacles. Restricting open-source models would trigger First Amendment challenges, and the US has limited leverage over overseas operations — Apple already deploys Alibaba's generative AI platform for iPhones sold in China.

Beijing is mirroring the approach. Chinese regulators have urged domestic firms against using homegrown AI abroad, and the Ministry of Industry and Information Technology claims Anthropic's Claude Code contains a backdoor posing a "serious threat." Both sides are weaponizing security reviews to fracture the global AI supply chain.

The practical result: multinational companies now navigate contradictory compliance regimes where model choice depends on jurisdiction, not merit. Expect more fragmentation as governments treat AI infrastructure like telecommunications gear — strategic, sovereign, and increasingly walled off.