White House Plans to Require U.S. Companies to Publicly Disclose Use of Chinese AI
nashnova research
The White House is drafting rules that would force U.S. companies to publicly disclose their use of Chinese AI models — framed as 'discourage, not ban' — betting on reputational pressure rather than outright prohibition.
What exactly would the rule require?
The Treasury Department and the White House Office of Science and Technology Policy are co-developing a rule: companies must publicly disclose on their websites any use of Chinese AI models.
The key framing is "discourage, not ban" — no direct prohibition on calling Chinese models.
This means → regulators are choosing transparency as the weapon, pushing both the choice and the reputational risk onto the companies themselves.
How deep has Chinese AI already penetrated?
OpenRouter platform data: seven of the top ten most-used models this week are Chinese.
Open-source models — predominantly Chinese — now handle over 80% of compute on the platform; a separate November report found Chinese labs account for 89% of U.S. enterprise open-source AI spending.
In plain terms = U.S. companies talk about "AI sovereignty," but most of their open-source compute already runs on Chinese models — penetration is far deeper than the public realizes.
Who gets hit first?
Reuters flagged a specific case: Airbnb uses Alibaba's Qwen model to power customer service.
If the rule lands, companies like Airbnb would have to disclose this on their websites — turning a back-end technical choice into public information.
This means → the question for companies shifts from "can we use it?" to "are we willing to let everyone know we do?"
What else is the U.S. alleging?
The U.S. government and AI labs accuse Chinese model developers of "distillation" — training on the outputs of another lab's model to accelerate their own progress.
Reportedly, Moonshot and DeepSeek used output data from OpenAI's GPT-6 Astra to train their latest products.
This reflects a widening concern: the U.S. worry has escalated from "can Chinese AI compete?" to "how was Chinese AI built?" — intellectual-property disputes are now layering on top of supply-chain risk.
Will this actually work?
The core logic: mandatory disclosure bans nothing by itself — it bets that reputational pressure will push companies to switch suppliers voluntarily.
In plain terms = the government doesn't pull the trigger on a ban; it pins a "uses Chinese AI" label on companies and lets the market and public opinion do the pressuring.
The real unknown: will companies proactively drop Chinese models to protect their image, or will they wait passively until regulators escalate to an outright ban?
市场有风险,内容仅供研究参考,不构成投资建议。
