Yesterday’s news — that Chinese AI models now account for 30 to 46 percent of US enterprise API traffic — apparently made waves in Washington. CNBC reports today that the US Congress is investigating the growing adoption of Chinese AI models by American companies.
Who’s investigating, and who’s targeted?
The House Committee on Homeland Security and the House Select Committee on China have launched a joint investigation. First step: letters to Cursor and Airbnb, asking about their use of and risk exposure to Chinese AI models.
Coinbase CEO Brian Armstrong publicly posted that his company uses GLM-5.2 (Z.ai) and Kimi 2.7 (Moonshot). The State Department warned CNBC that Chinese AI models are “designed to advance Beijing’s narratives, censor dissent, and reflect CCP ideology and values.”
Why companies are switching anyway
The reason is simple: money. Chinese open-source models cost 60 to 90 percent less than comparable models from Anthropic and OpenAI. For tasks that don’t need frontier performance — summarization, code completion, data extraction, support drafts — a model that costs 90 percent less and performs within 5 to 10 percentage points is the rational choice.
Vercel CEO Guillermo Rauch put it plainly: “Price is doing the work here.”
The dilemma
Banning Chinese open-source models is practically impossible. The weights are openly available on Hugging Face and GitHub — anyone can download them. Even if the US government prohibited companies from using them, enforcement would be nearly impossible.
My take
What’s happening here is a replay of the TikTok pattern: a Chinese product becomes popular because it’s better or cheaper than the US alternative. Then comes the political backlash. The difference from TikTok: open-source model weights can’t be removed from the internet with a ban. Washington will need to get creative — or accept that the AI market is going global just like the rest of the tech industry.
Sources: