Minerva co-founder joins debate as US weighs Chinese AI disclosure rules
The Trump administration is reportedly considering making US companies publicly reveal their use of Chinese AI models, a move that could reshape the economics of AI adoption
The Trump administration is reportedly considering a requirement that would make US companies publicly disclose when they use Chinese AI models.
Nothing has been finalized, and the discussions were still ongoing as of November 2026. The AI industry has not settled what customers will actually pay for intelligence, and this policy lands right in the middle of that argument.
Bloomberg spoke with a co-founder of Minerva about how companies are using Chinese AI models as Washington’s scrutiny intensifies. The conversation reflects a question many technology buyers now face: is a cheaper model worth the political baggage that may come with it?
What Washington is weighing
The proposal under discussion centers on transparency rather than prohibition. US firms would be asked to say openly whether Chinese-built models sit inside their products and operations.
That distinction matters. A disclosure rule would not stop anyone from using a model. It would make the choice visible to customers, regulators and shareholders, which may be enough to change behavior on its own.
Airbnb has been cited as one example of a company using Alibaba’s Qwen models. The reference shows the issue is not limited to obscure startups experimenting in a garage. Recognizable US consumer brands are already part of the picture.
Open-source models are a big part of why that debate exists. When a model’s underlying weights are freely available, a company can download it, run it on its own infrastructure and avoid paying per-query fees to a US provider.
The crypto and markets news you actually need—every morning.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The security case building against Chinese models
In September 2026, US intelligence agencies issued an advisory accusing 6 Chinese labs of carrying out industrial-scale distillation of US AI capabilities, a practice the advisory said dates back to late 2024.
Distillation works a bit like a student who never attends the lecture but carefully studies the top student’s answers. A smaller model is trained on the outputs of a larger, more capable one, absorbing much of its skill without repeating the expensive work of building it from scratch.
Then came a November 2026 report pointing to hardware concerns. It identified what it described as “fingerprints” of Nvidia Blackwell GPUs in recent models from DeepSeek, Moonshot AI, MiniMax and Ant Group.
Blackwell chips are export-restricted to China. If those models were trained using hardware that was not supposed to reach Chinese labs, the finding points to potential violations of existing US export controls.
How we got here
A Chinese company released a model that cost far less than leading US systems yet appeared to perform nearly as well as the cutting edge. That release punctured a comfortable assumption. The prevailing view had been that frontier AI required enormous budgets and the best American chips, giving US labs a durable lead. DeepSeek suggested the gap might be narrower, and cheaper to close, than many had believed.
Since then, Chinese models have kept gaining ground as capable, lower-cost alternatives to US systems.
What this means for AI companies and investors
For companies like Airbnb, the calculation shifts from purely technical to partly political. A model choice that once lived in an engineering document could end up in a public filing, where customers, lawmakers and investors can all weigh in.
The chip findings carry their own risks. Scrutiny over Chinese models tied to export-controlled Nvidia hardware could lead to sanctions or tighter rules, squeezing Chinese developers further.
Several things are worth watching from here. First, whether the disclosure idea moves from internal discussion to a formal requirement, and what form that takes. Second, how companies already using Chinese models respond, whether by switching, staying quiet or defending their choices. Third, whether the distillation and Blackwell allegations lead to concrete enforcement action against the named labs.
Disclosure: This article was edited by John Chen. For more information on how we create and review content, see our Editorial Policy.

