Skip to content
AnalysisAI

Name and Shame for Chinese AI: Trump Officials Weigh a Disclosure Plan

Treasury and White House science officials have discussed making US companies disclose on their websites when they use Chinese AI. No rule exists yet.

Michael Kratsios, seated in a white armchair, holds a microphone and gestures while speaking onstage at CES 2026 beside Kinsey Fabrizio.
Michael Kratsios, director of the White House Office of Science and Technology Policy, at a CES 2026 fireside chat on artificial intelligence on January 7, 2026. File photograph. His office is one of two that discussed the disclosure idea, according to Reuters.Xuthoria / Wikimedia Commons · Adapted photograph · CC BY-SA 4.0

The Trump administration is weighing whether to require American companies to state on their own websites that they use artificial intelligence models built in China, Reuters reported on Wednesday. The goal is to discourage those models through public pressure, without new restrictions that could anger the industry or Beijing. Reuters described the approach as “name and shame.”

Officials at the Treasury Department and the White House Office of Science and Technology Policy (OSTP) have discussed the idea, according to three people who spoke to Reuters. The report said it was unclear whether the proposal was gaining traction or would become formal rulemaking, and it described no timeline. The White House and Treasury did not respond to Reuters’ requests for comment.

View Reuters's post on X

A disclosure rule could reach companies such as Airbnb, whose AI customer-service agent runs on Qwen, a family of models from China’s Alibaba. In an October 2025 interview with Bloomberg, Airbnb’s chief executive, Brian Chesky, said the company preferred Qwen in certain situations because it was “fast and cheap,” words that House investigators later quoted back to the company.

In a statement to Reuters, Airbnb said its AI activity “runs overwhelmingly on U.S.-origin models.” It said it uses “only a limited number of China-origin models, all of which are open-source,” which “run only through approved U.S.-based service providers, keeping data and operations separate and protected.”

Why companies download Chinese models

The models at issue are open-weight. Their core trained components, called weights, are published so anyone can download, run and modify them. A company can host such a model on its own servers or with an American cloud provider instead of sending customer data to a service in China. That is a different arrangement from using a Chinese chatbot app.

Chinese open models have spread through American business because they cost less than tools from OpenAI and Anthropic and perform well enough for many jobs, Reuters reported. Citing OpenRouter, a service that routes requests among AI models, it said Chinese models were the three most used by volume of text processed in October. China hawks fear that Beijing is using models such as Alibaba’s Qwen, Moonshot AI’s Kimi and DeepSeek to make American companies dependent on Chinese AI, according to the report.

Chesky has argued that the security fear misreads the technology. “We are not providing data to any Chinese companies,” he told Bloomberg, which published his remarks on May 20. “An open-source model does not have access to data. It doesn’t work that way.”

Rep. John Moolenaar, in a gray suit and glasses, listens to Joey Siu in front of an American flag.
Rep. John Moolenaar, who leads the House select committee on China, with the human rights activist Joey Siu after a committee roundtable on Hong Kong on May 23, 2024. File photograph.Adam Jackman / House Select Committee on the CCP · Public domain · Adapted photograph.

The case for a public label

Congress has pressed the same companies for months. On April 29, Rep. Andrew Garbarino of New York, who leads the House Homeland Security Committee, and Rep. John Moolenaar of Michigan, who leads the House select committee on China, opened an investigation with letters to Airbnb and to Anysphere, maker of the Cursor coding tool, whose Composer 2 model was built on a version of Kimi. “The Committees have serious concerns about the national security and data-security implications of that approach,” their letter to Airbnb said. In late July, the chairmen said they had held productive discussions with both companies, and they wrote to DoorDash, whose co-founder said some tasks route through Kimi.

“Disclosure is an important step in managing the potential risks posed by the diffusion of Chinese open-weight models in the United States,” Moolenaar told Reuters. He said the risks remain even when American companies run those models on their own infrastructure. That challenges the main defense that companies such as Airbnb have offered.

Some of the worry concerns what the models say, not only where data travels. In October, CBS News reported on research suggesting Qwen may give government-friendly answers to controversial questions, including questions about Hong Kong.

CBS News’ October 2, 2026 report on new research into how Qwen answers questions about Hong Kong and other controversial topics.

Chinese AI model Qwen gives alarming answers to questions about Hong Kong, controversial topics on YouTube

The talks also fit a wider confrontation. In April, Michael Kratsios, the White House science office’s director, accused China of stealing American AI work on an industrial scale. In July, Treasury Secretary Scott Bessent said the administration could sanction Chinese AI companies found to have stolen intellectual property through distillation, a method of training one model on another’s answers. Reuters said revelations about Nvidia co-founder Jensen Huang’s affair with an alleged Chinese spy have renewed calls in Washington for a tougher line. So far, US officials have stopped short of restricting Chinese AI firms.

Scott Bessent, seated on a sofa between Marco Rubio and Pete Hegseth, gestures with both hands while speaking.
Treasury Secretary Scott Bessent speaks during a meeting with President Ferdinand Marcos Jr. of the Philippines in the Oval Office on July 22, 2025. File photograph.Daniel Torok / The White House · Public domain · Adapted photograph.

The case against a disclosure requirement

Nathan Lambert, an AI researcher who co-founded and directs Trillium Labs and previously co-led the Olmo open-model project at the Allen Institute for AI, has argued for stronger American open models. Replying on X to a post summarizing the report, he wrote that the plan “would be an own goal nonsense.”

View Nathan Lambert's post on X

Reuters listed American open-weight options, including Meta’s Muse Glimmer, Nvidia’s Nemotron and Reflection AI’s Beam. A website label would not settle whether those models match Chinese rivals on price and quality for a given task. It would also reopen an older argument about what happens once a powerful model is released for download.

China’s embassy in Washington said politicizing trade and technology would “only stifle global AI advances and serve no one’s interests.” It added: “We will resolutely safeguard Chinese businesses’ lawful rights and interests.” Alibaba did not respond to Reuters, and Moonshot and DoorDash did not immediately respond.

What we expect next

Our prediction is that the idea remains a proposal in the near term. Reuters described no timeline, and the administration’s broader resistance to new regulation pulls against a formal rule. An executive order, a published rulemaking notice or a bill requiring disclosure would strengthen the case that it is advancing. A public denial from the White House or Treasury would weaken it.

We also expect more companies to explain where their Chinese models run, as Airbnb has, before any rule requires it. Statements naming the model, the host and the data controls would show that public pressure is already working. OpenRouter data showing Chinese models losing share to American open-weight releases would be the clearest sign of changed behavior. Whether a label would move customers at all remains unknown.

Common questions

Is the US requiring companies to disclose their use of Chinese AI?

No. As of November 18, 2026, Reuters reported only that the Trump administration is weighing a proposal; no rule has been issued and no timeline was reported.

Which officials discussed the Chinese AI disclosure proposal?

Officials at the Treasury Department and the White House Office of Science and Technology Policy discussed it, according to three people who spoke to Reuters.

What would US companies have to disclose?

Under the idea Reuters described on November 18, 2026, companies would state publicly on their websites that they use AI models developed in China.

Why would the Trump administration use disclosure instead of banning Chinese AI?

The aim is to discourage reliance on Chinese models through public pressure without new restrictions that could anger the industry or Beijing, according to Reuters.

What is an open-weight AI model?

It is a model whose trained weights, its core learned parameters, are published for anyone to download, run and modify. A company can host one on its own servers or with a US cloud provider.

Which Chinese AI models are US companies using?

Reuters named Alibaba’s Qwen, Moonshot AI’s Kimi and DeepSeek. Citing OpenRouter, Reuters said on November 18, 2026 that Chinese models were the three most used by token volume in October.

Why are China hawks worried about Chinese open-weight AI models?

China hawks fear Beijing is using cheap models such as Qwen, Kimi and DeepSeek to make American companies dependent on Chinese AI, Reuters reported. Lawmakers have also raised data-security concerns, and research reported by CBS News suggests Qwen may give government-friendly answers.

Does Airbnb use Chinese AI models?

Yes. Reuters reports that Airbnb’s AI customer-service agent uses Alibaba’s Qwen, and Airbnb says it uses “only a limited number of China-origin models, all of which are open-source.”

Does running a Chinese open-weight model send data to China?

Not directly, according to Airbnb’s Brian Chesky, because a downloaded model can run on US servers. Rep. John Moolenaar says risks remain even when companies host the models themselves.

Which companies has Congress questioned about Chinese AI models?

Airbnb, Anysphere (maker of Cursor) and DoorDash. Two House committee chairmen wrote to Airbnb and Anysphere on April 29, 2026, and to DoorDash in late July 2026.

What did Rep. John Moolenaar say about disclosure?

Rep. John Moolenaar, who leads the House select committee on China, told Reuters in a statement published November 18, 2026 that “Disclosure is an important step in managing the potential risks posed by the diffusion of Chinese open-weight models in the United States.”

How did China respond to the US proposal on Chinese AI disclosure?

China’s embassy in Washington said politicizing trade and technology would “only stifle global AI advances and serve no one’s interests.”

Are there American open-weight AI models?

Yes. Reuters named Meta’s Muse Glimmer, Nvidia’s Nemotron and Reflection AI’s Beam as American open-weight models.

Has the US restricted Chinese AI companies?

Reuters reported on November 18, 2026 that US officials have so far stopped short of restricting Chinese AI firms. It said Treasury Secretary Scott Bessent said in July 2026 that the administration could sanction Chinese AI companies found to have stolen intellectual property through distillation.

Sources

All news
Donald Trump sits with senior national security officials around a conference table.
Strategy

The Iran Attack Scenario Washington Spent a Week Discussing

Jensen Huang stands in the audience at a White House AI summit in Washington.
Business

Trump Is Open to a Huang Return at Nvidia. Bessent Wants an Outsider.

Donald Trump and Xi Jinping shake hands at Gimhae International Airport in Busan, South Korea.
Strategy

Trump, Xi and Putin Will Sit Down Together for the First Time

Have data to license or need a source?

Tell us what your company has or what your team needs.

Talk to SnowRock