
Illustration
Developments3 min read
Report: US weighs forcing companies to disclose Chinese AI model use
The Trump administration is weighing a requirement for US companies to publicly disclose which Chinese AI models they use, Reuters reported on 18 November, citing three people familiar with the discussions. Companies would post the information on their own websites; the plan would not ban models such as Qwen, Kimi or DeepSeek, according to the report.
Key facts
- Reuters reported on 18 November 2026 that the Trump administration is weighing a rule requiring firms to disclose Chinese AI model use on their websites.
- Treasury and the White House OSTP discussed the plan, according to Reuters; it would not ban the models, and whether it would go through formal rulemaking is unclear.
- Airbnb uses a Qwen model for customer service and Cursor's Composer 2 was built on a version of Kimi, Reuters reported.
Researched and written with AI, to Luigi's brief, positions and editorial rules. Automatically checked against the sources. How I work →
The Treasury Department and the White House Office of Science and Technology Policy (OSTP) have discussed the idea, according to the report. It would apply to companies such as Airbnb, which uses a model from Alibaba's Qwen family for its AI customer service agent. It was not clear whether the idea was gaining traction or whether it would go through formal rulemaking, Reuters said. The White House, Treasury and Alibaba did not respond to Reuters' requests for comment.
Reuters described the approach as a name-and-shame campaign meant to curb reliance on Chinese models without new AI rules that could anger industry or Beijing. So far, the US has not restricted or punished any Chinese AI company, according to the report.
Which companies use Chinese AI models
Airbnb told Reuters its AI activity runs overwhelmingly on US-origin models. It said it uses only a limited number of China-origin models, all open source and run only through approved US-based providers. Republican John Moolenaar, chair of the House Select Committee on the Chinese Communist Party, welcomed the deliberations and said risks remain even when US companies run the models on their own infrastructure.
Moolenaar and Andrew Garbarino, chair of the House Homeland Security Committee, launched an investigation into US companies' use of Chinese AI this year, Reuters reported. Letters went to Airbnb, to DoorDash, whose co-founder said the company routes some internal tasks through Moonshot AI's Kimi, and to Anysphere, maker of the Cursor coding tool, whose Composer 2 model was built on a version of Kimi.
Price is the draw. Chinese open-weight models are cheaper than proprietary tools from OpenAI and Anthropic and good enough for many tasks, Reuters wrote; by token usage, the three most popular models on the OpenRouter marketplace in October were Chinese. Open models overall recently accounted for 71.7% of OpenRouter tokens, according to Hugging Face, as covered in the Hugging Face and OpenRouter partnership story. Alibaba now also offers Qwen4-Max hosted in a Frankfurt data centre. The Chinese embassy in Washington said Beijing opposes politicising trade and technology issues.
Analysis
For now this is only under discussion inside the administration, and it concerns US companies. The underlying question, though, applies to any business that buys AI through software: which model actually runs inside the tool. The Cursor case shows the answer is not on the label. According to Reuters, anyone using Composer 2 works with a Kimi-based model without having chosen it. Hugging Face's new router, set to “:cheapest”, also picks the provider on price, the company says.
A US disclosure rule would make that gap visible to customers, partners and procurement teams, including European firms that work with US groups. A short model list for every AI tool in use is therefore worth keeping now: vendor, model, origin, where it runs and who is responsible internally. It is the same unglamorous work as data protection and approvals, and it decides whether a team can answer a customer question quickly or only after weeks. Someone inside the company should own the choice of AI suppliers. The guide to AI marketing and the law covers the legal side of using AI in marketing, and the Cursor breach via Fireworks showed how quickly data ends up with suppliers.
Got AI news? Submit news →
FAQ
Do US companies already have to disclose Chinese AI models?
No. According to Reuters on 18 November 2026, the administration is only weighing the requirement; whether it comes and whether it goes through formal rulemaking is unclear.
Which Chinese AI models are meant?
Reuters names Alibaba's Qwen, Moonshot AI's Kimi and DeepSeek. The plan would not ban these models, according to the report.
Source
Reuters: Trump administration weighs forcing US firms to disclose Chinese AI use
More sources
- Hugging Face – Anteil offener Modelle an den OpenRouter-Tokens (71,7 Prozent), siehe eigener Artikel vom 18.11.2026.



