Technology

Artificial Intelligence News: Trump Eyes China AI Disclosure

The artificial intelligence news from Washington on Wednesday pointed to a new front in the Trump administration's contest with Beijing over advanced technology. Officials are weighing a plan that would require American companies to publicly disclose when they use Chinese-made AI models, according to three people familiar with the discussions who spoke to Reuters. The idea would not ban the technology or punish the companies that rely on it. Instead, it would put the fact of that reliance in public view, on the companies' own websites, in the hope that embarrassment and customer pressure do the work that a formal restriction might not.

The plan is still only an idea, and it is far from certain to become policy. The Treasury Department and the White House Office of Science and Technology Policy have discussed it, but it was not immediately clear whether the proposal has real momentum inside the administration, or whether it would ever be written into a formal rule. What is clear is the direction of travel: after years in which Washington's main weapon against Chinese technology was the export control, officials are now looking for softer tools that discourage American companies from building on Chinese AI without starting a fight with industry or with Beijing. The disclosure plan is the latest of those tools, and it arrives in a week when the administration's whole China technology policy is under the brightest spotlight it has faced all year.

How the Disclosure Plan Would Work

Under the plan as described to Reuters, companies that use Chinese AI models would have to say so publicly, most likely by posting the information on their own websites. The example cited by the people familiar with the discussions was Airbnb, the home-rental company, which uses Qwen, an AI model built by the Chinese technology group Alibaba, to power its AI customer service agent. A traveller asking the company's automated helper a question is, in effect, talking to a Chinese-built model — and under the proposal, Airbnb would have to tell the public that. The White House, the Treasury Department, Alibaba and the Chinese Embassy in Washington did not respond to requests for comment, and Airbnb declined to comment.

Advertisement

The appeal of the idea, for the officials considering it, is that it costs the government almost nothing and asks the companies to do the hard part themselves. A ban on Chinese models would invite lawsuits, complaints from American businesses that depend on cheap and capable tools, and a diplomatic row with Beijing at a moment when the president is trying to keep the trading relationship steady. A disclosure rule, by contrast, works through reputation. A company that has to admit on its own website that its customer service runs on a Chinese model may think twice about the choice — and its customers, investors and rivals will all be able to see the answer. Officials described the approach as a "name and shame" campaign, pursued alongside a broader reluctance to regulate artificial intelligence directly.

That reluctance matters. The administration has made a point of its light-touch approach to AI, arguing that heavy regulation would slow American innovation. A disclosure requirement fits that philosophy: it does not tell a company what technology it may use, it only asks the company to be open about the choice. Whether that is enough to change behaviour is an open question, and it is one the officials behind the plan have not answered. The proposal's supporters would say that sunlight is a powerful incentive; its critics would say that most customers will never read the disclosure page, and that companies will keep using whatever tool is cheapest and works best.

AI News: Why Chinese Models Are Winning US Customers

The reason the administration is worried at all is that Chinese AI models are no longer a curiosity in the American market. They are, by one widely watched measure, the most popular models in the world. Chinese AI models were the three most used by token volume in October on OpenRouter, a marketplace platform that tracks how much work different models do. Tokens are the units of data that AI models process, so the ranking is a rough measure of real-world usage — and in October, the top three spots all belonged to Chinese models.

Advertisement

The appeal is straightforward. Chinese open-weight models are cheaper than the proprietary tools sold by the leading American labs, OpenAI and Anthropic, and they perform well enough for most business uses that the price gap is hard to ignore. Open-weight models make their core components publicly available for anyone to download and adapt, which means a company can take a powerful Chinese model, run it on its own systems, tune it for its own needs, and pay far less than it would for an American alternative. For a business choosing the engine for a customer service agent, a coding assistant or an internal search tool, that combination of price and performance is difficult to turn down.

The names at the centre of the debate are now familiar in the industry. Qwen, Alibaba's model family, powers Airbnb's customer service agent. Kimi, built by the Chinese start-up Moonshot AI, sits behind the Composer 2 coding model in Cursor, the popular AI coding tool made by the American company Anysphere. DeepSeek, whose models stunned the industry when they appeared, remains one of the most widely used open-weight systems in the world. The concern among China hawks in Washington is not only that American companies are using these tools, but that the tools are designed to create dependence: once a company's products, workflows and data are built around a Chinese model, switching away becomes expensive and disruptive, and Beijing gains a lasting foothold in the American technology stack.

Cybersecurity News: Distillation, Data and the Hawks' Case

The hawks' case goes beyond dependence. Treasury Secretary Scott Bessent said in July that the administration could sanction Chinese AI companies if they were found guilty of stealing American technology through distillation — the process of training a smaller, cheaper model by feeding it the output of a larger one, as a way of cutting the cost of building a powerful new tool. The accusation behind that warning is that Chinese labs have used distillation to copy the work of American companies, getting the benefit of billions of dollars of American research without paying for it.

Advertisement

Bessent's warning echoed an accusation made in April by Michael Kratsios, the director of the White House Office of Science and Technology Policy, who said China was stealing the intellectual property of American artificial intelligence labs on what he called an "industrial scale." The claim set off a State Department-led effort to raise the issue with foreign governments, and it has since become a staple of the administration's argument that Chinese AI is not merely a competitor but a product built, in part, on stolen American work.

The most striking example of the problem, for the administration's critics, involves the American government itself. Reuters reported in September that a government website had deployed an AI model from Alibaba to let users search proposed federal regulations — even though the FBI had accused the model's developer of copying technology from Anthropic, one of the leading American labs. The episode showed how far Chinese models have spread, and how hard it would be to unwind their use: if a federal agency can end up running an Alibaba model without apparently noticing the provenance, a private company can hardly be blamed for doing the same. It also showed why the disclosure idea appeals to officials who have stopped short of bans: a rule that forces the use into the open would at least make it visible.

American open-weight models do exist, and the administration's supporters point to them as the alternative. Meta's Muse Glimmer, Nvidia's Nemotron and Reflection AI's Beam are all open-weight systems built by American companies, and they offer businesses a way to get the benefits of an open model without relying on a Chinese one. The question is whether they can match the Chinese models on price and performance — and, for now, the usage figures suggest they are losing that contest.

Advertisement

Congress Is Already Asking the Same Questions

The administration is not the first part of the government to notice the trend. Congress has been asking the same questions for months. Republican Representatives Andrew Garbarino of New York, who chairs the House Homeland Security Committee, and John Moolenaar of Michigan, who chairs the House select committee on China, opened an investigation this year into American companies' use of Chinese AI. The two chairmen sent letters to Airbnb, to Anysphere and to DoorDash, the food delivery company, whose co-founder has said the company routes some tasks in an internal system through Kimi.

The letters were a sign that the concern crosses party lines and branches of government, and that the disclosure idea did not come out of nowhere. Lawmakers have been pressing companies to explain their use of Chinese models since well before the Treasury and White House discussions reported this week, and the congressional inquiry has already put the same names — Airbnb, Anysphere, DoorDash — on the record. Moonshot AI, DoorDash and SpaceX, the rocket company that bought Anysphere, did not respond to requests for comment on the Reuters report. The congressional investigation is still running, and its findings could give the disclosure proposal a push — or provide the evidence for a tougher approach.

Nvidia News: The Scandal Behind the Pressure

The disclosure debate is also inseparable from the scandal that has dominated Washington's technology conversation for the past week. Jensen Huang, the chief executive who built Nvidia into the most valuable company in the artificial intelligence boom, resigned on November 12 after revelations that he had extramarital affairs, including one with a woman American officials have identified as an alleged Chinese spy. The company's board has since ruled out bringing him back, and Colette Kress, the company's finance chief, is running the company on an interim basis.

Advertisement

The affair has turned a corporate scandal into a national security story, and it has intensified the pressure on the administration to show it is serious about Chinese technology threats. The revelations have drawn fresh attention to cases in which Nvidia's advanced AI chips were smuggled to China and went unpunished, and they have given the hawks in Congress and inside the administration a powerful argument that the existing controls are not working. Every new measure — the disclosure plan, the data center rule, the equipment curbs — is now read in Washington as part of the administration's answer to that charge.

Nvidia itself added a new fact to the picture on Tuesday. Kress told investors on the company's earnings call that Nvidia had sold $2.4 billion worth of licensed Hopper-generation data centre products to China in the quarter that ended on October 25, but that it had no current plans for further shipments, and that its guidance for the coming quarter included no licensed Hopper products at all. The Hopper family includes the H200, a powerful processor used to train and run advanced AI models. Washington had approved H200 sales to ten Chinese companies, including Alibaba, Tencent and ByteDance, the owner of TikTok, after the administration eased its licensing policy in January. Beijing has since restricted H200 imports, and Reuters reported earlier this month that around 600,000 of the chips were sitting in storage in Hong Kong, waiting for shipment to the mainland. Kress did not explain why Nvidia had stopped shipping the processors, or whether the halt would be permanent — but the semiconductor chip news of the week now has the industry's biggest name pulling back from China at the very moment Washington is debating how far to go.

Data Center News: The Parallel Crackdown on Chips

The disclosure plan is only one of several measures moving through the administration at the same time. On Monday, the Commerce Department issued a new rule requiring data centers outside the United States to obtain a government licence before selling significant computing power to firms headquartered in China, Russia, Iran and a small group of other high-risk countries. Officials say the rule closes a loophole that let Chinese firms rent the advanced Nvidia chips they are barred from buying outright — access that has helped China's artificial intelligence industry advance rapidly. The rule takes effect in 90 days, and it landed in the middle of an open argument inside the administration over how tough the chip controls should be.

Advertisement

The argument spilled into public view on Monday, when the president described the disagreement between his two cabinet secretaries in unusually candid terms, saying one wanted to be friendly with China and the other wanted a confrontation. Treasury Secretary Scott Bessent has pushed for months for a hard line on chip exports; Commerce Secretary Howard Lutnick, whose department writes and enforces the export rules, has resisted the toughest measures. The president said both views were acceptable, and that he would make the decision himself. The data center news of the past two days has shown the hawks gaining ground: the rule was issued, and Reuters reported on Tuesday that the administration also plans to press ahead with a plan to block imports of Chinese data center equipment, despite lobbying from Amazon, Google, Meta, CoreWeave and Cisco. The companies warned White House officials that replacing the Chinese hardware — optical transceivers that move data through fibre-optic cables at the speed of light — could take as long as a decade.

What Happens Next

For now, the disclosure plan remains exactly that: a plan under discussion. It is not clear whether it has the support it would need inside the administration, and it is not clear whether it would be written into a formal rule even if it did. The officials who described it to Reuters were careful to present it as an idea being weighed, not a decision taken — and the administration has a recent history of floating tough-sounding measures on China and then pulling back.

The timing adds a layer of delicacy. Trump is in China this week, in the southern tech city of Shenzhen for the Asia-Pacific Economic Cooperation leaders' meeting, and he is due to sit down on Thursday with Xi Jinping and Vladimir Putin for the first three-way summit of the leaders of the world's three largest nuclear-armed states. A disclosure rule is designed, in part, to avoid angering Beijing — it names no company, bans nothing and fines nobody — but it is still a public step against Chinese technology, taken while the president is on Chinese soil. Whether the hawks or the pragmatists win the argument will become clearer after the summit, when the administration has to decide whether the plan moves from discussion to rulemaking. Until then, the message to American companies is plain enough: Washington is watching what AI they use, and it is considering making sure everyone else can see it too.

Advertisement

Source: Reuters

Advertisement

Leave a Reply

Your email address will not be published. Required fields are marked *

Advertisement
Advertisement
About Us  |  Contact Us  |  Privacy Policy  |  Disclaimer  |  Terms and Conditions