The US government is preparing to push back against American companies that have been quietly turning to Chinese-made AI models to power their operations. According to Engadget, a State Department spokesperson said the use of Chinese AI models “by US companies raises serious concerns,” adding that these tools are “designed to advance Beijing’s narratives, censor dissent and reflect CCP ideology and values.”
The concern is not just political. US companies have been drifting toward Chinese AI for a very practical reason: cost. CNBC reported that Coinbase CEO Brian Armstrong publicly shared that his company had been using two Chinese-built models, GLM 5.2 from Z.Ai and Kimi 2.7 from Moonshot. The CEO of startup Lindy also confirmed a switch to DeepSeek specifically to cut surging operational costs.
This is not a fringe trend. A Nikkei Asia report from just days ago flagged that major US companies, including Airbnb and Uber, have been racing to adopt Chinese AI models. That report also linked the surge to Anthropic suspending access to two of its models, Mythos 5 and Claude Fable 5, reportedly at the request of the US government, leaving some businesses scrambling for alternatives.
So why does this matter beyond corporate bean-counting? The worry in Washington is that Chinese AI tools, even when they work well, may come with built-in ideological guardrails or data risks that American companies are not fully accounting for. The State Department’s framing makes that concern explicit, though no hard evidence of active data misuse has been made public.
The bigger picture here is a deepening tech cold war. Both sides are now actively trying to keep the other’s AI out of their own corporate ecosystems. China’s Ministry of Industry and Information Technology has claimed that Anthropic’s Claude Code contains a backdoor that poses a “serious threat,” a claim Anthropic has not confirmed. Chinese authorities have also reportedly held talks with domestic companies about avoiding foreign AI tools.
What the US can actually do to limit corporate AI choices is a real question. Changing government procurement rules is straightforward. Telling private companies which AI they can run internally is far harder, and restricting open-source models could run into First Amendment arguments before it ever reached a courtroom. The geographic dimension adds another layer of complexity. Apple, for example, uses Alibaba’s generative AI platform for iPhones sold in China. Washington can hardly dictate what tools a company uses to comply with local laws in another country.
The likely path forward involves a mix of:
- Updated federal procurement rules that bar Chinese AI from government contracts and contractors
- Pressure campaigns and guidance aimed at nudging private companies away from Chinese models
- Potential export-style controls framed around national security, rather than a blanket ban
None of that is settled yet, and the legal and diplomatic complications are real. What is settled is that Chinese AI has found a genuine foothold in US corporate America, and Washington has decided that is a problem worth acting on. How far that action goes, and how fast, is the story to watch from here.
