White House accuses Moonshot AI of cloning US tech as Treasury threatens sanctions

Treasury threatens sanctions after White House claims Moonshot distilled Anthropic's Fable

U.S. Treasury Secretary Scott Bessent doubled down on his warnings to Chinese AI companies on Wednesday, saying that sanctions remain on the table after a White House official accused Moonshot of improperly distilling Anthropic's Fable model. Model distillation is a common AI training technique in which a smaller model learns from the outputs of a larger one. While this process can infringe on intellectual property rights, it's also widely used as a legitimate optimization method. "Open source is not open season on American IP," Bessent posted on X. "When [Chinese] firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table." Earlier this week, Bessent stated that the U.S. government would examine open source models from China for signs of intellectual property theft and impose sanctions if found. Bessent's latest remarks come hours after the White House's science and technology policy chief Michael Kratsios accused the China-based Moonshot of conducting large scale distillation against U.S. models. He alleged that Moonshot had acquired Nvidia's "GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models," raising questions about whether the firm violated U.S. export control rules. The GB300 servers are part of Nvidia's Blackwell generation, which are banned from being sold to Chinese companies. Some experts dispute the idea that Kimi K3 could have been developed primarily through distillation from Fable, which has only been publicly available since July 1. Moonshot released K3 last week as an open-weight model, and its advanced capabilities have called into question the underlying business models of leading U.S. AI labs, casting doubt on whether they can continue to justify the enormous capital requirements underpinning the frontier AI race. The episode has also intensified a broader debate in Washington over the influx of Chinese open models. Some, including former White House AI advisor and current OpenAI Head of Strategic Futures, Dean Ball, have argued that the U.S. should restrict or effectively ban the use of Chinese open-weight models to preserve America's technological advantage and mitigate potential national security risks. TechCrunch has reached out to Moonshot and the Treasury for comment.

China's AI models have Trump's AI world at war with itself

Kimi and other free models from China have again been seen as a wake-up call. But for what? This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Over the weekend, several current and former advisors to President Donald Trump on AI publicly lobbed insults at the country's leading AI companies. David Sacks, the president's AI and crypto "czar" until March, branded Anthropic's models as "lobotomized" and "woke." Emil Michael, a top Pentagon official, called OpenAI's new head of strategic futures a "supreme village idiot." It began because no one can agree on what to do about Kimi, a free, open source model that Chinese AI company Moonshot launched last week. It appears to rival the intelligence of models from OpenAI and Anthropic, which are very much not free. Kimi and other Chinese models like it pose a real problem for Trump. And they're dividing the top AI strategists in his orbit into factions. Every time a new smart, free model from China like Kimi gets released, US companies see less reason to fork out money to access models from Anthropic or OpenAI. Given that enthusiasm for these and other AI companies is driving an outsized share of economic growth, China's AI models create both economic and political problems for the president. They are "a threat for an administration that really doesn't want more economic bad news," Anton Leicht, a fellow at the Carnegie Endowment, wrote on X. They've already rattled US stocks. What is Trump to do? First, consider that this is all happening just a week after New York imposed the country's first state ban on new data centers. There is growing distrust of AI companies, and I imagine a not-insignificant share of Americans would have little sympathy for OpenAI or Anthropic as they fend off cheaper competitors, and would say it's not the government's job to protect their interests. On this point, they'd see a sliver of agreement (and really just a sliver) with David Sacks, who on July 19 criticized top AI companies that "want the government to eliminate their open source competition." He has also argued that Chinese AI models have become popular because they come with fewer restrictions on how people can use them (putting aside the built-in state censorship). Sacks, however, is out of a job. He no longer has a formal role advising Trump, and his position that more open AI is better has been largely replaced in the administration by one that sees a larger role for government intervention. The thinking behind this view is that because AI models have gotten strong enough to pose threats to national security, the government must control how they're used. This position has fueled the new White House review process that aims to vet AI models' security before they're released. Dean Ball, a former Trump AI advisor who now works for OpenAI, criticized it over the weekend as a "de facto licensing regime for frontier AI." Ball predicted Trump may solve his Chinese open source problem with a bit of soft power, perhaps by making US companies afraid to use models like Kimi. That drew a response from Michael, who, with Secretary of Defense Pete Hegseth, has been the agency's main liaison with AI companies. Michael called Ball the AI industry's "supreme village idiot," bristling at the suggestion that the government would quietly strong-arm companies rather than, as Michael put it, go through "the democratic process not some Deep State scheme." Left out of the conversation has been how a model like Kimi got so good in the first place. For much of the Biden administration and even the beginning of Trump's second administration, keeping China from getting top chips was a priority. Those export controls have loosened -- Trump made the controversial decision to allow Nvidia to sell more chips to China, in exchange for the US government taking a cut -- and the government has alleged that some chip smuggling has taken place. But China nonetheless has limited computing power, and it's not clear what chips the company behind Kimi used to train the model. It's possible that the process involved some distillation, a practice in which AI models are trained on the outputs of existing AI models. OpenAI and Anthropic have long complained that Chinese AI companies do this, and they have requested government help to put a stop to it. In April, they got it, when the Trump administration announced a series of efforts to curb the practice. But Kimi is out there and free, and it is nearly as good as the Anthropic model the US government deemed so powerful that it was briefly shut down because it threatened national security. The weekend's sparring suggests many in Trump's orbit see that as a wake-up call. But nobody can agree on what for.

Silicon Valley Is Completely Divided Over Chinese AI

A huge debate is brewing in Silicon Valley over the proliferation of Chinese-made artificial intelligence tools, particularly "open-weight" AI systems that, by some measures, can compete with or even outperform some of the best US models. My WIRED colleague Hugo Lowell has written about the Trump administration's internal debate on how to handle these Chinese models. Among AI companies in the Valley, the issue is proving even more divisive. A top concern in both DC and the Valley relates to distillation, in which a less capable AI model is trained on the outputs of a more powerful one. In June, Anthropic accused the Chinese tech giant Alibaba of illicitly stealing its IP through distillation attacks. Then, earlier this week, the White House said that it believes the Beijing-based Moonshot AI had developed its Kimi K3 model by distilling Anthropic's Fable 5 model. Another big concern is how quickly China's models are appearing and spreading. An open-weight AI model has its core components made public, so that it can be fine-tuned to suit a user's needs. But they don't have the kinds of guardrails on which Anthropic has been building its reputation. Yasir Atalan, deputy director and data fellow at the Center for International and Strategic Studies, points out that the main benefit of open-weight AI models is their speed of diffusion. They can spread especially easily "through Hugging Face, GitHub, cloud providers, local deployments, and third-party inference platforms," he writes. If you're Anthropic, and you've built a cult around safety and charge for access to your big expensive proprietary models, you have every reason to want to regulate this. But some Silicon Valley startups -- not the trillion-dollar ones like OpenAI and Anthropic -- really don't want the US government to put restrictions on these AI models. On Wednesday a group of over 200 startups called the Little Tech Association sent a letter to Michael Kratsios, science adviser to President Donald Trump, and US Commerce Secretary Howard Lutnick lobbying against an outright ban of open-weight AI models. The group, which includes famed startup incubator YCombinator, has argued in favor of certain safeguards but says that denying Americans access to AI models abroad would weaken US startups and create a monopoly among the AI giants. Bill Gurley, the legendary tech investor and longtime partner at Benchmark Capital, has publicly argued in favor of letting "the free market work." In a lengthy blog that offers a nice little history of open-source software, Gurley writes that open-weight models avoid lock-in, encourage true academic research, and are critical for capital-constrained startups. "Every AI startup, every solo developer, every two-person team building a product on top of AI infrastructure depends on having access to good models at affordable prices," Gurley says. Chamath Palihapitiya, one of the All-In podcast hosts, wrote on X that "tricking the US Government to protect frontier labs' business model by using a China boogeyman is a mistake...It is protecting the equity of 5,000 people who are investors in OAI and Ant at the sale of everyone else. This would be a terribly stupid decision." His cohost and fellow VC Jason Calacanis piled on. "Daddy Trump protect us!!!!" he wrote on X, with an alarming number of crying-laughing emoji. This stance from some of Silicon Valley's most ruthless capitalists might at first seem counterintuitive. Why let a foreign adversary's technology flourish in the US? It's as if the US is up just 1-0 in the AI World Cup, a slightly uncomfortable lead, and the crowd is chanting for the opposing team to get a free kick. The reason, as always, is money. In a sense, open-weight models are the "move fast and break things" of the AI era, if the new motto tacked on something like "and use a scalpel to fix it." Having access to open-source software allows startups to scale, scale, scale -- and deal with the consequences down the road. Meanwhile, the AI labs and hyperscalers that make proprietary platforms, like OpenAI, Anthropic, Google, Microsoft, Meta, and XAI, stand to benefit greatly if their systems remain protected and dominant. The bigger question that none of these companies seem to be asking is what best serves the 99 percent of us who don't have their financial future fully staked on advancing AI. Anthropic CEO Dario Amodei has warned repeatedly that open-weight LLMs present an untenable security risk, because they can be downloaded by anyone and tuned to malicious ends. That argument might sound hard to dispute, except that the recent Hugging Face hack happened to show the opposite. After an OpenAI model escaped containment and infiltrated the open-source platform, "our own forensic work was blocked by the guardrails of the hosted models we first tried," Hugging Face wrote on its blog. The company then turned to a Chinese open-weight model to help resolve the threat. The US government has much to weigh as it decides how to handle Chinese open-weight models. Thank goodness this administration doesn't seem to be at all financially motivated. This is an edition of Steven Levy's Backchannel newsletter. Read previous newsletters here.

US threatens sanctions against Chinese AI models over IP theft

On Tuesday, Treasury Secretary Scott Bessent said the U.S. would examine open source models from China for signs of intellectual property theft, threatening sanctions against Chinese AI companies if IP theft is established. "We've seen a lot of talk about open source models coming and threatening the large language models in the US," Bessent said on Fox Business Tuesday. "This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft." Bessent's comments were first reported by Bloomberg. The statement comes as Chinese models -- most recently Moonshot AI's Kimi K3 -- are gaining in capabilities and popularity, threatening to harm the business models of top American AI firms like OpenAI and Anthropic, as well as their abilities to raise more capital to continue developing frontier models. On Monday, Axios reported that the Trump administration is considering a wholesale ban on Chinese open source models, although others have disputed that reporting. AI companies have been warning for months against campaigns by foreign actors to copy their AI technology and redeploy it as open source. In April, the White House said it would work closely with AI firms to combat the theft. Sanctions from the U.S. against Chinese models would add to the growing list of strategies the government is attempting to maintain the lead in the AI race. After restricting China's access to advanced chips and tightening export controls, Washington is now signaling it may target the AI models themselves, a move that could mark a significant escalation in the technological competition between frontier labs and Chinese open-source alternatives. Model distillation is a technique that allows some of a larger model's capabilities to be translated into a smaller system that's easier to run -- but not everyone agrees that distilling another company's model constitutes theft. Earlier this month, Microsoft CEO Satya Nadella criticized large labs for making just this assumption: "While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation." AI labs' training practices continue to be a source of legal risk for the companies. Anthropic this week got the green light to start cutting authors checks as part of its $1.5 billion settlement after a judge ruled it had illegally downloaded and stored millions of copyrighted books to train its AI. Furthermore, some in the industry argue that distillation isn't the only reason China is catching up to U.S. AI companies. "We know distillation to be a very small factor in the ability to create good models, and it's a practice that everyone is doing, including companies in the U.S.," Hugging Face CEO Clem Delangue said on a recent episode of TechCrunch's Equity podcast. "If it were easy just to do distillation to get good at building AI models, there would be many other countries, including in the U.S., with much better open source AI. The reality is they have really, really good research teams in China...taking a much more open and collaborative approach to AI than in the U.S."

White House: China's Kimi K3 AI Model Cheated by Cloning US Tech

China's new AI model, Kimi K3, has been making waves as the latest threat to US dominance in artificial intelligence. But according to the White House, the Chinese model is merely an unauthorized knock-off based on US tech. On Wednesday, Michael Kratsios, director of the White House Office of Science and Technology Policy, claimed that Kimi K3's developer, the Chinese startup Moonshot AI, created the model by effectively cheating and cloning the capabilities of Anthropic's cutting-edge Fable model. "We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," Kratsios tweeted. "To do this, they developed a sophisticated internal platform to conduct large-scale distillation against US models, allowing them to quickly switch between multiple methods of access to avoid detection." Distillation usually refers to training a smaller AI model to mimic the performance of a larger one. But Chinese companies have allegedly been using that process to extract and clone capabilities from top-performing US models. In February, Anthropic accused Moonshot, DeepSeek, and MiniMax -- all Chinese firms -- of using distillation to uncover the reasoning and coding capabilities of its Claude chatbot. In Moonshot's case, the Chinese startup allegedly used hundreds of fake accounts involved in 3.4 million exchanges with Anthropic's chatbot. Kratsios didn't elaborate on Moonshot's alleged distillation to create Kimi K3. But he also accused the Chinese startup of sourcing cutting-edge Nvidia Blackwell GPUs from Thailand, thus bypassing US export controls on the technology. Distillation isn't exactly illegal, though. Nor was any evidence given. Still, Kratsios appears to be setting the stage for a White House crackdown on Chinese AI models over national security concerns. "Large-scale, covert industrial distillation aimed at stealing proprietary US technology and undermining American research is unacceptable," he tweeted. Anthropic's Head of Public Policy, Sarah Heck, also chimed in, tweeting: "Illicit, adversarial distillation is IP theft and industrial espionage that supports adversary military and intelligence capabilities. It is a national challenge that creates serious national security risks for the United States and democratic allies." According to Axios, the Trump administration is considering banning US companies from hosting Chinese models unless their security can be guaranteed. Treasury Secretary Scott Bessent also publicly said this week: "If we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft." Moonshot AI didn't immediately respond to a request for comment. But the arrival of Kimi K3 has also set off a debate about the US's closed-source approach to AI versus China, which has been pursuing open-weight AI models that can cost less to use. Tech investor Chamath Palihapitiya tweeted his own criticism, writing: "Tricking the US Government to protect frontier labs' business model by using a China boogeyman is a mistake. It is protecting the equity of 5,000 people who are investors in [OpenAI] and [Anthropic] at the sale of everyone else. This would be a terribly stupid decision. Let the market sort this out!"

Senior White House official claims China's K3 model stolen from Anthropic

Donald Trump's Assistant for Science and Technology, Michael Kratsios, has accused China's Moonshot AI of creating its head-turning Kimi K3 model distilling Anthropic's Fable. Kimi K3 is 2.8-trillion-parameter open-weights model of such impressive quality that its mere existence suggests Chinese AI researchers aren't far behind their US rivals. Moonshot AI released it on July 16, and not long afterwards the value of US AI stocks sank as investors worried the model could damage their businesses. Kratsios used a Xeet to allege that Kimi K3 is the result of distillation - a technique that involves bulk queries of one model to train another - rather than innovation. "We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," he wrote, adding the assertion that the Chinese company "developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection." Kratsios also accused the company of gaining access to servers packing Nvidia's GB300 accelerator - a model the US does not allow to be sold in China - and of accessing GB300s running in Thailand. Kratsios added his view that the USA "strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models." He also noted that AI distillation can be a legitimate technique when "used to create smaller, more efficient models." "However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable." US Treasury Secretary Scott Bessent weighed in to the matter with a Xeet of his own that opens "We support open-source AI and the innovation it unlocks." If you feel like there's a "but..." coming, you're right. "But open source is not open season on American IP," Bessent wrote. "When [People's Republic of China] PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table." The US has already sanctioned just about every major player in China's tech ecosystem, but that hasn't prevented the nation from growing enormous and sophisticated tech companies. Nor have those sanctions stopped Chinese firms from creating "evasion routes" to secure access to banned tech through illicit means, or by using the grey market. One evasion route is renting GPU farms outside China to run AI workloads, to access hardware that's not available in the Middle Kingdom. Anthropic accused Moonshot AI of distillation in February 2026 and said its Chinese peers DeepSeek and MiniMax also used the technique. The US and other nations have often accused China of industrial espionage. Beijing always denies such allegations. Whatever means China uses to develop technology, think tank the 2025 Australian Strategic Policy Institute's Critical Technology Tracker rates the Middle Kingdom as the leader in 66 of the 74 technologies it rates. ®

As AI grows more powerful, a US-China feud threatens safety efforts

BEIJING, July 24 (Reuters) - U.S. threats to sanction Chinese AI developers over alleged intellectual property theft and export-control violations risk unravelling efforts to establish a bilateral dialogue on AI safety, analysts say, just as increasingly powerful models raise concerns over serious security breaches. U.S. officials on Wednesday accused Chinese AI lab Moonshot of distilling its Kimi K3 model from Anthropic's advanced Fable 5 model, prompting Treasury Secretary Scott Bessent to warn of potential sanctions. Distillation is the process of training AI models using output from more advanced ones to lower the costs of training a powerful new AI tool. The Commerce Department's Bureau of Industry and Security is also investigating whether Chinese firms like Moonshot are accessing advanced U.S. chips illegally to train their models. A Moonshot spokesperson did not respond to Reuters' request for comment. The dispute exposes a growing tension at the heart of the U.S.-China AI race. Both sides increasingly view frontier AI as a strategic asset and a security risk, yet experts argue cooperation on safety standards is becoming more urgent as powerful models could be misused for cyberattacks ⁠or other harmful applications. The accusations, coupled with reports Washington is weighing restrictions on Chinese open-weight models, could provoke tit-for-tat measures and jeopardise a planned U.S.-China AI dialogue in September, analysts say. The tensions come atop years of U.S. export controls designed to curb China's access to cutting-edge chips, measures that have increasingly dominated trade talks between the two countries. "Depending on the number of Chinese companies targeted, (and) the nature of the punitive actions taken... the retaliation has the potential to scuttle both the AI dialogue and the September 24 meeting between Presidents Trump and Xi," said Paul Triolo, a partner at DGA-Albright Stonebridge Group. Beijing meanwhile is considering whether to restrict overseas users' access to its models as a retaliatory measure, Reuters previously reported. CLOSED AND OPEN-SOURCE SECURITY RISKS New York-based Hugging Face last week used a Chinese model, Z.ai's GLM-5.2, to contain a cyberattack by a rogue OpenAI agent that escaped during safety testing, since American closed models' safety guardrails were ironically too strong. As Chinese and U.S. frontier models achieve recursive self-improvement (RSI), where systems can autonomously enhance their own capabilities, researchers say both countries have an incentive to cooperate on safety standards before a more serious incident occurs. Industry experts are also highlighting the risks posed by open-weight models, which can be downloaded, modified and redistributed with little oversight, limiting the effectiveness of software-based export controls. "Godfather of AI" ⁠Yoshua Bengio warned at China's flagship AI forum last week that the deployment and sharing decisions of open-weight models are irreversible, and their safeguards are easier to remove. "The logical thing to do is to find a good evaluation of these models, share the models that are not too dangerous, and not share those above the threshold of risk," he said via videolink. RESOURCE, GEOPOLITICAL CONSTRAINTS Chinese AI models must undergo government safety and content reviews before release, though current rules do not cover post-release modifications. Unlike U.S. giants OpenAI and Anthropic, many Chinese AI labs lack the computing resources for extensive safety training and instead focus on improving model capabilities, industry observers say. "It's plausible China reconsiders allowing open-weight releases for frontier models down the ⁠line, but the bar for doing so is high," said Kristy Loke, a MATS research fellow studying China's AI governance. Some AI safety researchers argue that frontier models in both the U.S. and China should face stricter pre-release testing and independent third-party evaluations. In the U.S., firms can voluntarily submit models for testing by the state-funded Center for AI Standards and Innovation (CAISI), though some lawmakers have called for mandatory federal reviews. "In an ideal world, the two countries will come together to work on ⁠safer models... agree to build common standards around pre-release testing and set red lines for the most advanced open models," said Loke. DIVIDED RESPONSE ON CHINESE AI MODELS Some U.S. government and AI industry officials remain split on how to respond to Chinese models, which account for about 60% of token usage by U.S. companies on the OpenRouter platform. OpenAI and Anthropic have lobbied Washington against lower-cost Chinese models, arguing they could undermine their business ⁠models. OpenAI lead strategist Dean Ball wrote on X that the Trump administration could "create large amounts of regulatory risk around the use of open-weight Chinese models" to curb their adoption by U.S. firms. White House AI adviser David Sacks countered that leading U.S. labs "want the government to eliminate their open-source competition" and argued for fair competition, adding in another post that the "Kimi Panic needs to stop". "As long as we don't sabotage ourselves with unnecessary rules, the U.S. will continue to win," he said. Reporting by Laurie Chen in Beijing; Additional reporting by Samuel Shen in Shanghai; Editing by Eduardo Baptista and Shri Navaratnam Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Disrupted * Data Privacy * Intellectual Property Laurie Chen Thomson Reuters Laurie Chen is a China Correspondent at Reuters in Beijing, whose coverage focuses on the nexus of frontier technology, strategic emerging industries and geopolitics. She has reported on China for almost a decade, having previously covered China's government, defence, security and foreign policy. She has broken multiple global scoops on U.S.-China relations and the trade war 2.0, elite Chinese politics and diplomacy. She is particularly interested in Chinese frontier AI, tech and industrial policy, semiconductor supply chains, robotics, aerospace and grand strategy.

Moonshot AI accessed Nvidia's chips despite Chinese export ban, White House official says

It comes as the race for AI supremacy between the U.S. and China intensifies. A White House official has accused Chinese AI company Moonshot of accessing Nvidia's advanced chips, despite export controls banning them from doing so. Moonshot unveiled Kimi K3 on Friday. It closes the performance gap with frontier models including Anthropic's Fable 5 and OpenAI's GPT 5.6 Sol, even surpassing them on some benchmarks. It's the largest open source model so far, with 2.8 trillion parameters, referring to the size of its neural network. The Chinese company "acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models," Michael Kratsios, director of the White House Office of Science and Technology Policy, said in a post on X on Wednesday. Nvidia, Moonshot AI, the White House and the U.K. embassy of the People's Republic of China have been approached for comment. Nvidia's most advanced AI chips are under export restrictions to China, though some less capable semiconductors are allowed to be shipped to the country. Nvidia's GB300 chips are one generation behind its frontier Vera Rubin systems and are still considered cutting edge. The release came as the race for AI supremacy between the U.S. and China intensifies. Alongside limiting access to crucial AI hardware, lawmakers in America are also considering how to curb the growing adoption of Chinese AI models by Western companies. ByteDance was reportedly assembling computing power outside of China as it looked to boost AI efforts, according to a Wall Street Journal report in March. ByteDance has been approached for comment. The U.S. is looking to close export control loopholes that allow advanced AI chips to be accessed by Chinese companies. The Remote Access Security Act (RASA), a proposed bipartisan legislation, seeks to expand U.S. export controls to include the remote cloud-based access of critical hardware and software. It passed the House of Representatives in January and is awaiting action in the Senate. "It is well known that China is using American compute both within the country and abroad," said Keegan McBride, director of science and technology at think tank the Tony Blair Institute for Global Change. "It demonstrates just how far ahead the US is on compute," he added. "American compute is the foundation for the AI race today." While the U.S. reportedly cleared 10 Chinese firms to buy Nvidia's H200 chips in May, earlier this month a top U.S. trade official said "very few" had been shipped to China and Hong Kong. With access to the leading AI chips severely limited, Chinese tech companies have increasingly sought homegrown alternatives. White House official Kratsios also said "we have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," echoing Anthropic's claims earlier this year of "industrial-scale" campaigns to extract information from its model. Through distillation, smaller AI models can mimic the performance of larger, pre-trained models by extracting knowledge from the better-trained model. A spokesperson for the U.K. embassy of the People's Republic of China previously told CNBC that the country "opposes baseless allegations and malicious smears against its AI development." Treasury Secretary Scott Bessent threatened sanctions on Chinese companies that commit distillation attacks on Wednesday. "When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table," he said in a post on X. Choose CNBC as your preferred source on Google and never miss a moment from the most trusted name in business news.

China's Moonshot AI stole from Anthropic, Trump tech advisor says

A White House advisor has accused China's Moonshot AI of a "large scale" effort to steal the capabilities of top US artificial intelligence (AI) models. US President Donald Trump's Science and Technology advisor Michael Kratsios said Moonshot AI carried out the campaign through what is known as distillation - when a weaker AI model extracts answers from a stronger one. Moonshot also gained access to restricted cutting-edge Nvidia servers to train its models, Kratsios said in a social post on Wednesday. The BBC has contacted Moonshot, Anthropic, the White House and Nvidia for comment. Kratsios said on X that the US government has information that Moonshot AI "distilled" capabilities from Anthropic's Fable AI for the development of its K3 model. Kimi K3 gained attention around the world after it was unveiled last week, with many believing it to have narrowed the gap between Western and Chinese AI models. Moonshot said its K3 model is able to rival top US technology. Kratsios' allegations come just a day after Treasury Secretary Scott Bessent said the US would examine whether Chinese AI models have stolen the capabilities from American rivals. "We've seen a lot of talk about open-source models coming and threatening the large language models in the US," Bessent told Fox Business on Tuesday. "If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them," he added. The increased scrutiny by the US of Chinese AI companies also comes as Trump is expected to meet his Chinese counterpart Xi Jinping in September.

Startups urge Trump not to ban Chinese open-weight AI

Nearly 200 US startups have told the Trump administration not to cut off access to Chinese open-weight AI, warning a ban would kill hundreds of them and hand the market to Anthropic and OpenAI. It is the first time Silicon Valley's little guys have organised against the industry's giants, and it has opened a rare rift. Silicon Valley does not usually gang up on its own. This week it did. Nearly 200 US startups have urged the Trump administration not to block American access to Chinese open-weight AI models. A ban, they warn, would gut the next wave of US companies. The plea, first reported by Politico, came in letters from the newly formed Little Tech Association. The signatories include Y Combinator and the privacy company Proton. On Wednesday they wrote to President Trump, Commerce Secretary Howard Lutnick, and the White House science chief Michael Kratsios. Their message was blunt. Do not reach for a broad ban. Use targeted safeguards instead. Why startups want the Chinese models Open-weight models are those whose weights are published, so anyone can download and run them. Moonshot's Kimi K3 and Alibaba's models are cheap, capable, and free to build on. Cash-strapped founders lean on them rather than pay for credits from Anthropic or OpenAI. A ban, the group argues, would not stop the models spreading. It would just hand the market to a few giants. "There'll be hundreds of companies that instantly die," said Suhail Doshi, founder of the startup Particle and an association member. "It's great for Anthropic. We're all going to have to spend money on Anthropic." A rare rift with Anthropic That is the sharp edge of the story. The fight has split the AI industry. Anthropic, with OpenAI, has pushed hardest for curbs on Chinese models, citing security. The startups say bans would do little for security and plenty of harm to competition. The Information reported that the two labs now stand almost alone in the industry. The Little Tech Association wants precision. "A scalpel rather than a sledgehammer," its executive director Harry Godfrey said. The aim, he added, is the lightest touch that addresses real security concerns without raising costs or blocking access. Washington turns up the heat The revolt follows weeks of escalation. It began after Moonshot released Kimi K3, a model that beat Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol on coding tests. Reports then suggested the administration might ban Chinese models outright. Officials have kept up the pressure. Treasury Secretary Scott Bessent said on Fox Business that the US could "sanction" firms that steal American intellectual property. Kratsios went further. He alleged Moonshot had distilled Anthropic's Fable model and used banned Nvidia GB300 servers to train K3. Moonshot has not responded. Separately, Commerce's export-control arm has opened a formal inquiry into whether Chinese firms are getting restricted US chips. For now, the ban is only talk. A White House official called reports of one "baseless speculation", and said a blanket ban was not seriously discussed. But the rhetoric is not cooling, and the split it has exposed is real. The giants want walls. The startups want the door left open. And in Shanghai, Xi Jinping has been pitching cheap, open AI to the rest of the world.

US Treasury Chief Threatens Sanctions on Chinese AI Labs Over 'IP Theft' Concerns

U.S. Treasury Secretary Scott Bessent has "the ability to sanction," and he is not afraid to use it. In an interview with Fox Business on Tuesday, Trump's Treasury chief said that Chinese open-source AI models should expect increased scrutiny from American authorities. Over the last few months, American frontier AI labs and the Trump administration have accused Chinese open-source providers of IP theft related to advanced AI models developed in the U.S. "We've seen a lot of talk about open source models coming and threatening the large language models in the US," Bessent said. "This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them." The fight over models China and the United States have been locked in a contest for global AI domination, and it has been heating up as of late. The U.S. is still the leader in artificial intelligence, holding a commanding position not just on the frontier AI models through industry leaders like OpenAI and Anthropic, but also on AI chip technology through Nvidia, AMD, and Intel. But Chinese models are quickly bridging that gap. The first signs of this came in January 2025 in the form of DeepSeek's R1, which was as good at some common AI tasks as OpenAI's then-best model, and which didn't require the latest -- and most expensive -- Nvidia GPUs to run, making it significantly more cost efficient. Its release led to a trillion-dollar market freakout that drove the Trump administration to impose export bans and overall stricter trade restrictions on China. Then last week, Alibaba-backed Chinese AI startup Moonshot unveiled Kimi K3, its latest soon-to-be-open-weight model that ranks close to or above some of the most advanced closed-weight offerings from OpenAI and Anthropic on several benchmarks. The release and its incredible online reception challenged the long-standing assumption that Chinese AI labs were still several months behind American competitors. According to independent testing from Artificial Analysis and Arena.ai, Kimi K3's capabilities are competitive with both Anthropic's Fable 5, which was released last month, and OpenAI's GPT-5.6, which was released just a week prior. Back in February, though, Anthropic accused three Chinese AI companies -- Moonshot, DeepSeek, and Minimax -- of violating Anthropic's terms of service and regional access restrictions to "illicitly" extract the capabilities of its Claude model and use that to improve their own. Around the same time, OpenAI sent a memo to the House of Representatives, accusing DeepSeek of a similar tactic, claiming that it is trying to "free-ride on the capabilities developed by OpenAI and other U.S. frontier labs." Then, last month, Anthropic accused Chinese tech giant Alibaba of the similar behavior. Theft or standard-practice? This "illicit" "theft" process is by no means new, and is actually fairly common in the AI industry. The practice is called distillation, and put very simply, it's when one "teacher" model's output is used to train another "student" model. But in a memo from April, Trump's chief science and technology adviser deemed the China-led "industrial-scale" version of the practice "adversarial," and vowed to crack down on it by helping U.S. AI labs identify and thwart attempts at such distillation. On Tuesday, Bessent said that watermarks of American LLMs were found on many Chinese AI offerings, adding that the administration would be looking into the matter in "the coming days or weeks." Tensions rise ahead of U.S.-China AI talks Also on Tuesday, Reuters reported that the United States and China would have their first bilateral talks regarding artificial intelligence. Similar working-level talks between the two governments had taken place under the Biden administration, but this is the first such meeting of Trump's second term. Many of the details for these long-anticipated talks are undetermined, according to the report citing unnamed sources familiar with the matter, except that it will take place in September and be led by Bessent on the U.S. side. Bessent Will likely bring up the AI distillation concerns in the talks, if they take place, while the Chinese authorities are expected to raise their concerns regarding Mythos, Anthropic's scary new AI model that can allegedly hack into even the most secure systems. The Trump administration has only allowed Anthropic to share the model with a handful of companies and government agencies around the world, and Chinese companies and agencies almost certainly didn't make the cut considering Washington's national security concerns with sharing advanced AI technologies with Beijing. In a key speech at China's top AI conference last week, Chinese President Xi Jinping voiced his disapproval of the United States' approach, calling for open AI cooperation between nations. "We should seize this rare, historic opportunity to encourage open source, openness, collaboration, and sharing," Xi said. "In the meantime, we should jointly oppose overstretching the national security concept in the field of AI and placing one country's security over that of others."

US accuses China's Moonshot of stealing from Anthropic's Fable for latest AI model

WASHINGTON, July 22 (Reuters) - China's Moonshot AI stole from Anthropic's most sophisticated AI large language model, Fable, to help create its latest K3 release and has acquired advanced Nvidia AI chips, top Trump administration tech official Michael Kratsios said on Wednesday. "We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," he said in a post on X. "Large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable." U.S. Treasury Secretary Scott Bessent said on Wednesday in a separate post on X that he was considering adding Moonshot AI to a trade blacklist and imposing sanctions. The Trump administration's accusations will likely ramp up tensions over technology issues between Beijing and Washington as firms from both countries introduce increasingly powerful AI models that have broad economic implications and potential military uses. Moonshot did ⁠not immediately respond to a request for comment. Spokesperson Liu Chang of the Chinese embassy in Washington called Kratsios' comments "entirely unfounded," adding that China respects intellectual property protections. "Relevant individuals in the United States should respect the facts, discard prejudice and stop smearing and discrediting China's achievements in the development of its artificial intelligence industry," he said. Anthropic public policy executive Sarah Heck said in a post on X that Chinese theft of U.S. models "creates serious national security risks for the United States." Kratsios also said Moonshot AI had acquired servers containing powerful Nvidia (NVDA.O), opens new tab AI chips called GB300s and had used them in Thailand, "likely to train its AI models." A senior Trump administration official said in February that Chinese AI startup DeepSeek's latest AI model was trained on Nvidia's most advanced AI chip, in what could represent a violation of U.S. export controls. The U.S. State Department ordered a global push in April to bring attention to what it says are widespread efforts by Chinese ⁠companies, including Moonshot AI, to steal intellectual property from U.S. artificial intelligence labs, according to a diplomatic cable seen by Reuters. Distillation is the process of training smaller AI models using output from larger, more expensive ones as part of an effort to lower the costs of training a powerful new AI tool. In February, Anthropic said three Chinese artificial intelligence companies used its Claude model to improperly obtain capabilities to improve their own models. DeepSeek, Moonshot and MiniMax created more than 16 million interactions with Claude using roughly ⁠24,000 fake accounts, in violation of Anthropic's terms of service and regional access restrictions, the company said. OpenAI has warned U.S. lawmakers that DeepSeek was targeting the ChatGPT maker and the nation's leading AI companies to replicate models and use them for its own training, The April cable from the State Department said, "AI models developed from ⁠surreptitious, unauthorized distillation campaigns enable foreign actors to release products that appear to perform comparably on select benchmarks at a fraction of the cost but do not replicate the full performance of the original system." Chinese AI startup Moonshot on Friday unveiled Kimi K3, a 2.8 trillion-parameter model that it said is the ⁠world's largest open-weight AI system and delivers performance approaching U.S. giant Anthropic's frontier Fable model. The launch, which comes a month after Anthropic's Fable and Mythos models were abruptly withdrawn by the U.S. government due to security concerns, underscores how quickly China's open AI ecosystem is narrowing the gap with the most advanced U.S. systems. Reporting by Bhargav Acharya and Chris Sanders; Writing by Doina Chiacu; Editing by Michelle Nichols, Mark Porter and Daniel Wallis Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * United States * Securities Enforcement * Intellectual Property Courtney Rozen Thomson Reuters Courtney Rozen reports on the world's largest technology companies from Washington, D.C., focusing on the relationship between the tech industry and the U.S. government. She reported on DOGE and the federal workforce during the first year of U.S. President Donald Trump's second term. Prior to joining Reuters, she was a White House correspondent at Bloomberg Government. She graduated from American University with a master's degree in journalism. Contact: [email protected]

Bessent says U.S. could sanction China over AI model 'theft'

U.S. Treasury Secretary Scott Bessent on Tuesday said the Trump administration will look into whether Chinese artificial intelligence models have been distilled from American models, stating that the government does not support "IP theft." Chinese open-weight models are gaining steam against leading offerings from American companies like OpenAI and Anthropic, sparking concerns from tech executives and government officials about the durability of the U.S.' lead in the AI race. Moonshot AI, a Chinese startup, released a new model called Kimi K3 earlier this month that outperforms those companies across some industry benchmarks. "If we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," Bessent told Fox Business on Tuesday. Bessent said the technical term for this theft is called distillation, which is an AI training method where a small, less capable model is built using outputs from an existing, stronger model. Anthropic sent a letter to the U.S. Senate Committee on Banking, Housing, and Urban Affairs last month alleging that the Chinese tech company Alibaba had carried out the "the largest known distillation attack" against it to date. "We are finding watermarks of our U.S. large language models on many of the Chinese models, and that's unacceptable," Bessent said Tuesday. "We're going to be looking at that in the coming days or weeks." The U.S. and China are planning to hold talks around AI in September, according to a report from Reuters on Tuesday. Bessent will represent the U.S. during the discussions, the report said.

White House: Moonshot used banned Nvidia chips for Kimi K3

The accusation came in a social-media post from Michael Kratsios, the White House science director. Moonshot, he said, reached banned Nvidia chips in Thailand and distilled Anthropic's Fable model to build Kimi K3. Nvidia's own chief executive says the fix is to embrace Chinese AI, not ban it. The model that rattled Silicon Valley last week now has Washington's attention. A senior White House official has accused China's Moonshot AI of breaking US export controls to build Kimi K3. It reached restricted Nvidia chips, he said, and leaned on American models to do it. Kratsios directs the White House Office of Science and Technology Policy. He made the claim in a post on X, Bloomberg reported. Moonshot "acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models," he said. That breached both US export rules and the labs' terms of service. The GB300 is part of Nvidia's Blackwell line, which the US bars from sale to Chinese firms. The accusation follows months of official unease. Trump officials have been examining whether firms used a loophole to ship advanced chips to Chinese subsidiaries abroad. A May memo reaffirmed the ban on selling Blackwells to Chinese companies. The post went further on the software side. Moonshot distilled Anthropic's Fable model to build K3, Kratsios said. It ran a "sophisticated internal platform" for "large scale distillation against US models," switching access methods to avoid detection. Distillation trains a cheaper model on a rival's outputs. Kratsios drew a line: legitimate distillation is fair, covert theft of US technology is not. Nvidia and Moonshot did not respond to requests for comment. Why Kimi K3 matters Kimi K3 is the reason any of this landed. The open-weight model, released last week, quickly climbed the leaderboards. Moonshot says it trails only Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 on overall capability. Its arrival set off a rout in tech stocks, echoing DeepSeek a year earlier, and a wave of US alarm. A fight already running hot Kratsios was not alone. Treasury Secretary Scott Bessent told Fox Business that Washington could sanction Chinese labs for intellectual-property theft. The complaints echo long-running claims from OpenAI and Anthropic, which has accused Moonshot of distilling its models before. "This administration supports open source models, but what we do not support is IP theft," Bessent said, as Bloomberg reported. He later warned, in a post on X, that sanctions and entity-list designations "will be on the table." Nvidia's dissent As the worry spread, Nvidia chief executive Jensen Huang took the other side. He called open models like Kimi "excellent" and urged the US to embrace them, not ban them, Fortune reported. His case: shutting China out only pushes it to build its own stack. Z.AI has already done so, with a data centre running only Chinese chips. The stakes The timing bites. Moonshot is racing to close a funding round at up to $50bn. A Hong Kong listing could follow this year. Targeted US sanctions would not ban the model. Anyone can already download it. But they could keep American money away. Kratsios published no evidence for his claims. The model itself is not in doubt. What has shifted is the argument: no longer whether China can compete, but how it got there.

The US is Concerned Over 'IP Theft' by Chinese AI Labs

U.S. Treasury Secretary Scott Bessent threatened to impose sanctions on Chinese AI firms deemed to have stolen intellectual property from American technology companies. "This administration supports open source models, but what we do not support is IP theft," Bessent said on Fox Business yesterday (Tuesday). "If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft." It comes as Chinese AI models threaten to upend the tech industry in which the United States is the clear leader. Moonshot AI released Kimi K3 this week, which performs well despite not having access to the state-of-the-art computer chips that American firms have. According to TechCrunch, American companies have been warning the White House for months that foreign firms might copy their AI technology and release it as open source. Earlier this year, Anthropic accused Moonshot and two other Chinese AI firms, DeepSeek and MiniMax, of "illicitly" extracting Claude's capabilities. This practice is known as distillation, and it is reportedly common practice in the AI industry. As Gizmodo puts it, it's when the "teacher" model's output is used to train the "student" model. Pot Calling the Kettle Black Of course, a lot of people in the creative industries will see the irony when they read about AI firms complaining that their IP is being stolen. Microsoft CEO Satya Nadella called out this hypocritical behavior earlier this month in a tweet aimed at AI companies. "While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation," Nadella wrote. Bessent also noted that watermarks of American LLMs have been discovered on Chinese AI models. That is reminiscent of when the Getty Images watermark could be reproduced on pictures generated from Stability AI's model, Stable Diffusion. Bessent adds the Trump administration will be looking into the matter in "the coming days or weeks." Axios reported this week that the government is mulling a total ban on Chinese open source models. Image creditsHeader photo licensed via Depositphotos.

Headaches for Silicon Valley as China chips away at the US's lead in the AI race

Google's AI struggles scream trouble as new Chinese models (again) throw US tech dominance into question Hello, I'm Blake Montgomery, writing to you after a double-header feature of Christopher Nolan's The Odyssey and the World Cup final. What a great Sunday. Today in tech, we're discussing how China is chipping away at the US's lead in the AI race and how Silicon Valley's workers are taking action to protect their jobs from AI. As Google falters on AI, China eats into the US's lead Google is struggling with the latest version of its Gemini model A Chinese company, Moonshot, put out a free, surprisingly capable model last week China's new models yet again thrown the US dominance in AI into question Google is months behind in its launch schedule for the new version of its flagship AI model, Gemini 3.5. According to Bloomberg, Google is attempting to improve Gemini's coding abilities to catch up to Claude, the Anthropic bot that is widely considered the best AI coding tool. OpenAI released a new model at the start of July; Google had been expected to do the same in May. News of the delay sent Google's shares sliding as investors questioned whether it could maintain its top position in the AI race. A refreshed and newly capable Gemini might also have ameliorated the damage to the US's reputation as the leader in AI this week - but didn't. As Google's internal struggles became public, the Chinese startup Moonshot debuted an unexpectedly strong model, sending shivers through Silicon Valley and Washington DC alike. Moonshot's Kimi K3 model featured surprisingly capable coding, particularly in a niche called "front-end coding". "This may be the single biggest release of the year," Anastasios Angelopoulos, the CEO of AI testing company Arena, said. By Sunday night, Moonshot couldn't keep up with the demand for its latest model. "Kimi K3 has received far more love than we expected, and our GPUs are feeling it. Over the past 48 hours, demand has pushed close to the limits of our current capacity," the company posted on X. Unlike ChatGPT, Claude and Gemini, Kimi K3 is free to use and open-weight, meaning anyone can download it and run it locally on their own computers, as its core components are released to the public. It's been the general strategy of Chinese AI companies: open-source and open-weight models, beating US AI companies on price as Chinese manufacturers do, and mirroring the broader trade war of the US and China. DeepSeek inspired a major US market panic last year. Alibaba previewed its latest model, Qwen3.8 Max, on Sunday, showing off a similar number of parameters as Kimi K3. The company plans to publish the model's weights soon, per Bloomberg. The strategy has been made possible by Beijing, which subsidizes computing resources and power consumption for its domestic AI startups, allowing them to release their AI models for free. The basic logic goes that if these startups can't compete on capability with US models, they can compete on cost and undercut the US parent companies, which need to earn a profit eventually. It's all causing major headaches in Silicon Valley. There are no longtime incumbents in the AI business to whom customers will default. If those buyers, especially businesses who need lots of generated code, can turn to cheap or free alternatives, there might be little need to spend big on brand-name bots. The South China Morning Post, which is itself owned by the Alibaba Group, reported last week: "As US AI costs soar, global businesses pivot to China's low-cost, open-weight models." In the US, nerves are rattling. Hawkish Trump administration officials are rekindling their efforts to block any US access to Chinese AI models via de facto bans, commerce department rules or an executive order, Axios reported Monday. OpenAI's policy head for advanced AI, former Donald Trump adviser Dean Ball, wrote that Beijing's open-weight approach to AI would lead to "full AI communism" with government-run datacenters powering AI as a public good. He denounced that future. OpenAI has been sounding the alarm about Chinese models for years, which reads as part political disagreement and part business strategy, since Chinese open-weight models threaten OpenAI's projected profits. "I would guess that the Trump Administration will at some point realize that their best strategy here would be to create large amounts of regulatory risk around the use of open-weight Chinese models," Ball wrote. There is no alignment between the US's leading AI companies and the government players in the same industry. Emil Michael, the US Pentagon's chief technology officer, responded to Ball: "Every industry/ecosystem has its supreme village idiot. Dean Ball is that for AI." As Beijing presents a unified front, the White House's response is growing more scattered by the day. Trump himself had talked until recently of complete deregulation and never restraining any AI investment, research or releases. By contrast, the White House is restricting access - even allies' access - to Anthropic and OpenAI's advanced cybersecurity models. Even the personnel are changing. Chris Fall, the director of the Trump administration's Center for AI Standards and Innovation, is resigning just three months into his tenure, Axios reported Monday. Not all of the AI crowd is against these models' incursion on US dominance. Bill Gurley, a longtime venture capitalist best known for his early backing of Uber, wrote of China's AI models in the Washington Post: "They aren't a security threat - they're what competition looks like." China's advanced AI does pose a risk to the global business of AI, which is, as yet, still a race without a clear winner. Cheap Chinese AI models may trump their expensive US counterparts. The entire US economy is tied tight to the AI industry, raising the stakes of the Chinese models' capabilities. Undercutting the US's AI frontrunners' business could unsettle the entire US stock market even more so than it already has. That competition is set to play out in the future. In the present, though, China's recent releases have thrown the politics of AI in the US into chaos. That reads as the bigger win. Datacenters face global political blowback Silicon Valley's workers are worried about what AI is doing to their jobs Employees are taking some of the US's first collective actions to protect their jobs from AI Last week, Meta workers filed a lawsuit against their former employer alleging that it automated their firing via algorithmic decision-making. Meanwhile, thousands of Google's workers petitioned their CEO for layoff protections as the company shifts its spending from employees to AI infrastructure. "Make no mistake: this is a company that is enjoying massive, unprecedented success," Parul Koul, a Google software engineer and the president of the Alphabet Workers Union, said outside the company's California headquarters. "These layoffs and cuts are not difficult decisions, but simply profit being put over the people that make this company run." These Silicon Valley employees are making these moves against the backdrop of the AI boom, which is changing software engineering rapidly and causing unions to gain popularity. A Meta employee who was part of a recent union drive in the UK said: "It's about us having a voice in all of the operations of the company, from products to employee experience." Meta countered last week's lawsuit with an unusually direct statement: "These claims lack merit and are not based on facts. Workforce management and organizational decisions were and are made by people, not AI." Google did not issue a statement in response to last week's petition. We've seen Silicon Valley luminaries warn of futuristic apocalyptic scenarios, but we're seeing those worries come to life among the fearful rank-and-file employees, who are now taking action. We're witnessing how those workforces, who have a front-row seat to AI's development, respond to its effects on their employment. Read more from Reworked, a Guardian series about what's at stake as AI disrupts our jobs.

The secret Trump administration battle to fight Chinese AI

Why it matters: Parts of the administration have tried to implement de facto bans on foreign open-source models before, knowledgeable sources tell Axios. Last week's rise of Chinese model Kimi is reigniting those efforts. Pro-competition voices fear the implications. * U.S. companies are increasingly using open-source models from China because they're cheaper and, as seen by the advent of Kimi, just about as good as the domestic technology. Behind the scenes: The Commerce Department last year considered adding multiple Chinese AI labs to its "Entity List," which would effectively cut off U.S. access without a license, a source close to the administration told Axios. * The National Security Agency and White House Office of the National Cyber Director also considered putting out an advisory on Chinese AI lab threats last year, practically discouraging U.S. companies from using their tech, the source said. * The White House considered implementing an executive order saying U.S. companies could only host Chinese models if they could guarantee security and take liability if it were breached, the source added. * Commerce last summer also circulated draft rules within the administration leveraging its authorities to secure domestic supply chains to target Chinese open-source models, another source close to the administration said. Administration officials keen on keeping regulation from stifling innovation killed all of those efforts. * Key voices, such as former White House adviser Sriram Krishnan, have since left the premises and national security hawks have grown louder. * With that shift in personnel -- along with the rise of more powerful Chinese tech and fresh cybersecurity fears -- ban momentum is picking up again. For the record: Neither the White House nor the Commerce Department responded to requests for comment. The big picture: The administration wouldn't need to impose an outright ban to get U.S. companies to drop those Chinese platforms. * "What's actually happening is slower and more durable," one source familiar with government discussions said, citing procurement rules, Entity List threats and public pressure campaigns aimed at U.S. companies using Chinese models. * Instead of a ban, another source familiar with government discussions described a push to highlight potential backdoors and lack of security with Chinese models, and the governance issue that brings. * It's an offensive approach where the administration encourages a more innovative U.S. open-source ecosystem, the source added. Yes, but: Powerful U.S. alternatives are still limited at best compared to cheaper and more practical Chinese models. What they're saying: David Sacks, an outside White House AI adviser, wrote Sunday on X: "We are at a critical inflection point in AI policy. The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open-source competition." * Sacks has long warned against regulatory capture that would benefit the largest U.S. labs and stifle competition. * While not outright bans, the approaches under consideration would have a chilling effect on Chinese open-source tech and U.S. companies that rely on it, the source close to the administration said. * The source described leading AI labs or their allies approaching the administration every 3-5 months with an idea to ban open-source models. What to watch: Chinese leader Xi Jinping is feeling emboldened in the AI race. Pressure is mounting for the U.S. industry, open-source or not, to react. The bottom line: Anthropic and OpenAI have long been loud proponents of safety regimes that in some cases include licensing. Those pushes have been more abstract -- until now.

Scott Bessent warns China of sanctions over AI model theft

The Treasury secretary said the U.S. is finding watermarks of American large language models on Chinese AI systems Treasury Secretary Scott Bessent on Tuesday warned that the Trump administration could sanction Chinese artificial intelligence models if it finds they were built by stealing from American ones. "If we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," Bessent told Fox Business on Tuesday. Bessent identified the practice by its technical name -- distillation -- describing it as a method of AI training where the outputs of a large, powerful model are used to develop a smaller one. "We are finding watermarks of our U.S. large language models on many of the Chinese models, and that's unacceptable," he said. "We're going to be looking at that in the coming days or weeks." The warning follows a letter Anthropic sent last month to the U.S. Senate Committee on Banking, Housing, and Urban Affairs in which the AI company alleged that Alibaba, the Chinese technology giant, had executed what it called "the largest known distillation attack" against it to date, according to CNBC. Chinese open-weight AI models have been gaining ground against leading American offerings. Moonshot AI, a Chinese startup, released a model called Kimi K3 earlier this month that outperforms some American models across industry benchmarks, according to CNBC. Open-weight models release their final trained parameters for anyone to download, but the source code and data used to train them are not disclosed. The threat of sanctions comes as the two countries are planning their first official AI dialogue under President Donald Trump, with talks expected in September before Chinese President Xi Jinping's planned U.S. visit on September 24. Bessent is set to lead the U.S. delegation during those discussions. The Trump-Xi summit in May produced an agreement to establish intergovernmental AI talks, with Bessent saying at the time that meetings could begin within four to eight weeks. China has said it is seeking technical engagement on AI issues rather than political talks. Anthropic and OpenAI have each faced their own IP-related legal challenges. Anthropic reached a $1.5 billion settlement in September with a group of authors whose lawsuit accused the company of using books sourced from pirated databases without authorization. OpenAI and Microsoft $MSFT were sued by The New York Times in 2023 over claims that the two companies had trained their AI models on the publication's content without its consent.

It's Official: AI Execs Are Quaking in Their Boots

Can't-miss innovations from the bleeding edge of science and tech A Chinese open-weight AI model called Kimi K3, developed by Beijing-based firm Moonshot AI, has sent a shiver down the spines of AI tech executives. The powerful, 2.8 trillion-parameter model impressed with its competence, igniting a war with far more expensive alternatives being offered by the likes of OpenAI and Anthropic. Top executives at both companies are sounding alarm, the Wall Street Journal reports, watching as Chinese open-weight models are rapidly catching up to their most powerful proprietary models. As a result, they're begging the Trump administration to step in and protect them from the influx of cheaper alternatives, which could undermine their increasingly desperate attempts to attract new customers. Dean Ball, who joined OpenAI as the head of strategic futures after helping shape AI policy for the Trump administration, was seemingly rattled, arguing that allowing Chinese open-weight models to take over would result in "AI communism" in a controversial and widely disputed tweet. He also suggested the Trump administration would inject enough "fear, uncertainty, and doubt" through "regulatory risk" that would eventually deter hyperscalers from using Chinese AI. OpenAI CEO Sam Altman has also watched as China continues to catch up in the ongoing race, telling CNBC in February that the progress the country's tech sector had made was "remarkable" and "amazingly fast." Altman has also signaled that he's prepared to slash prices for the company's latest AI models. While he didn't directly mention the skyrocketing popularity of cheaper AI alternatives from China, Altman tweeted last week that OpenAI would be "happy to deliver" its latest flagship GPT-5.6 Sol AI at "one-quarter of the price." Anthropic CEO Dario Amodei also has a long track record of begging the US to stop selling AI chips to China, likening Nvidia selling its hardware to the country to "selling nuclear weapons to North Korea and then bragging that the missile casings are made by Boeing and so the US is 'winning,'" in a lengthy January essay. In June, Amodei called capable open-weight models a serious concern, arguing that the US government should have the power to block AI developers from deploying risky AIs. The incursion isn't just coming from China. As the WSJ notes, US-based AI labs are starting to switch to open-weight models. Just last week, former OpenAI exec Mira Murati's Thinking Machine Lab released its first model, which happens to be open-weight. The trend could put AI companies in a bind: how can they keep financing their enormous AI data center projects and advanced model development if potential customers start switching to heavily subsidized or free AI models that provide good-enough or even frontier capabilities? Early signs of an imminent exodus are certainly there. Moonlight AI was forced to pause new subscriptions to its blockbuster model just 48 hours after launch due to overwhelming demand, pushing its servers to capacity. It's a particularly precarious moment as frontier labs continue to hike up prices to start covering at least some of their unprecedented spending, despite growing fears over an AI bubble. Put simply, why shell out for Anthropic's Claude Code or OpenAI's Codex when there's a far cheaper and highly customizable option out there? Markets are seemingly painfully aware of the dynamic. The Nasdaq composite had a bruising couple of days following the announcement of Moonshot AI's Kimi K3 last week, only regaining some ground on Monday. It's a familiar refrain. In January 2025, DeepSeek's V3 open-weight large language model threw a tech-heavy stock market into chaos, offering the same capabilities as other alternatives while requiring only a fraction of the computing power. All eyes are now on the Trump administration. Anthropic has ramped up the pressure on individual states to regulate AI faster in light of little momentum on Capitol Hill. Critics, most notably Trump's AI czar David Sacks, have accused the company of using fear-mongering to shut out smaller AI labs, and possibly Chinese companies as well. Nobody knows whether federal regulations are on the horizon. A White House official told the WSJ that the Trump administration remains committed to promoting America's open-source ecosystem and strengthening its security. In other words, it's leaving all cards on the table for now -- a reality that has AI tech executives spooked as Chinese companies are eating their lunch.

The White House is dictating access to frontier AI models, shifting power from tech giants

The Trump administration has taken new steps to assert more control over the rollout of future artificial intelligence model releases by dictating which companies and entities are allowed access to the latest frontier models, two sources familiar with the matter told CNBC. Until now, that decision was in the hands of American AI giants. Both Anthropic and OpenAI have decided which companies and agencies have access to their most powerful models, and have often included major enterprise customers. Anthropic unveiled its most capable Mythos cybersecurity model to a handful of partners with Project Glasswing. OpenAI was asked by the administration to gate its recent GPT-5.6 release, and has a similar consortium called Daybreak for its cybersecurity model. A White House official told CNBC that it doesn't provide approvals for AI releases from private companies. The official said any engagements, testing or meetings with government experts are "voluntary" and that "decisions on timing and scope of releases rest entirely with the companies," referring CNBC to Trump's recent executive order. "The Administration continues to collaborate with all of America's frontier labs to strengthen the security of this technology without stifling innovation," they wrote. However, last month the Trump administration blocked Claude Mythos 5 and Fable 5 due to "national security concerns," reinstating access after weeks of intense negotiations with Anthropic. OpenAI last month said it would limit new AI models to "trusted partners" to comply with government requests.

Trump under pressure from all sides over Chinese AI surge

Washington (United States) (AFP) - US President Donald Trump is under pressure to respond to the increasing success of Chinese AI models that many companies prefer to the more powerful, but pricier, US versions from Anthropic or OpenAI. The issue came to the fore this week with the release of Kimi K3 by Moonshot, which White House officials say was developed by stealing the capabilities of Anthropic's most powerful model through a process called distillation. The method is widely used across the industry, but US officials say Chinese firms have deployed it at scale to illicitly copy proprietary American systems. Michael Kratsios, director of the White House Office of Science and Technology Policy, on Wednesday decried the practice and suggested that Moonshot had also circumvented US export curbs on powerful AI chips from Nvidia to build its products. The Commerce Department's Bureau of Industry and Security is formally investigating Chinese firms like Moonshot over their use of these chips, a spokesperson told The Information. Treasury Secretary Scott Bessent this week threatened sanctions against China, with reports the US is also weighing a ban or curbs on foreign-made open source models that are often built through distillation. Earlier this year Anthropic sent letters to US lawmakers warning about the practice, formally accusing Chinese firms Moonshot, DeepSeek and MiniMax of industrial-scale distillation of its Claude models. Illicit distillation "is IP theft and industrial espionage" and "a national challenge that creates serious national security risks for the United States and democratic allies," Sarah Heck, Anthropic's head of public policy, said on X. Open source, or open-weight, models cost less and use source code that programmers can customize -- something that is especially appreciated by startups and companies that do not want to be beholden to models from Anthropic, OpenAI or Google that cost money and must be taken as is. Even if just floated as an idea, the startup community has asked the White House to forgo any plan to block foreign-made open source models, saying it would just unfairly reward the US AI giants, thwart competition and stick entrepreneurs with high prices for AI. "Lobbyists are urging Washington to treat open-model AI as a security threat. In fact, it is something more familiar: proper competition that should be welcomed," wrote Bill Gurley, a former venture capitalist, in The Washington Post. On Wednesday, a group of 179 startups wrote a letter to the Trump administration urging it to think twice against any outright ban or strict curbs on foreign-made models. "Denying American startups access to models available abroad would stifle competition, entrench incumbents, and function as a tax on intelligence," the letter from the Little Tech Association said. 'Panic needs to stop' Some critics of limiting open source models believe that OpenAI and Anthropic are trying to save themselves by eliminating the emergence of this rival technology. The two AI giants are facing huge financial pressure ahead of expected IPOs in coming months, with Wall Street taking a more closer look than ever at their business performance before their expected public listings. "The Kimi Panic needs to stop," said David Sacks, the White House's former chief on AI policy, who still has the president's ear. "President Trump's light-touch regulatory approach is working...As long as we don't sabotage ourselves with unnecessary rules, the US will continue to win," he said on X. Weighing in the debate firmly for the Chinese models was Jensen Huang, the CEO of Nvidia, whose company makes AI's core infrastructure, whether for open source companies or US-made proprietary ones. American companies should "absolutely" be allowed to use Chinese AI models, he said to Axios. "These Chinese models are excellent. Open source models that are excellent should be used," he said.

New bill takes aim at Chinese AI companies accused of copying American tech

Sen. Adam Schiff, D-Calif., speaks at the Capitol on June 1.Anna Moneymaker / Getty Images file A bipartisan group of lawmakers said they were introducing legislation Thursday aimed at protecting American artificial intelligence companies like OpenAI and Anthropic from Chinese competitors. The bill, called the Collaboration on Adversarial Threats and Security Risks Act, treats Chinese alleged copycats as a security threat to the U.S. and would create an exemption to antitrust laws so AI companies and industry groups can coordinate on a response, according to a summary of the legislation from the lawmakers' offices. Antitrust laws generally limit intra-industry coordination on many subjects, on the theory that coordination may raise prices for consumers. The legislation is being introduced by Sens. Adam Schiff, D-Calif., and Jim Banks, R-Ind., and Reps. Bob Latta, R-Ohio, and George Whitesides, D-Calif. Lawmakers said the bill would allow AI companies and industry groups to share information related to certain security risks as long as they do so in good faith for the exclusive purpose of preventing, investigating or mitigating AI security risks. They could also coordinate -- for the exclusive purpose of reducing AI security risks -- by delaying or limiting the deployment of high-risk AI models, with required written notice before they take such actions, according to the summary of the legislation. The lawmakers said their antitrust exemption would include strict guardrails to keep the AI sector competitive. They said their legislation would also permit the attorney general to seek an injunction against organizations that abuse the exemption. It's part of a wider effort by the U.S. government to protect the fast-growing AI industry from mounting Chinese competition. In recent weeks, several Chinese AI companies, including Moonshot and Z.ai, released new AI systems that appear to match leading American AI capabilities -- a remarkable surge in Chinese AI performance. Treasury Secretary Scott Bessent threatened in an interview Tuesday with Fox Business to sanction "overseas models" for "stealing from our great companies." He repeated the threat Wednesday, saying on X that sanctions were "on the table." Michael Kratsios, director of the White House Office of Science and Technology Policy, separately accused Moonshot of "stealing proprietary U.S. technology" Wednesday on X. Moonshot did not respond to a request for comment. Anthropic, Google and OpenAI are increasingly complaining about what's known in the tech industry as "distillation": submitting thousands or even millions of queries to an advanced AI model and processing its outputs in an attempt to improve an inferior AI model and create, in effect, a less expensive knockoff. China has rejected allegations that it is stealing American trade secrets, calling them an unfounded smear. "China has consistently attached great importance to the protection of intellectual property and is committed to safeguarding, in a fair and impartial manner, the lawful rights and interests of both domestic and foreign IP holders," Liu Chang, a spokesperson for the Chinese Embassy in Washington, said in a statement Wednesday. "China's development is driven by its own sustained efforts and has also benefited from mutually beneficial international cooperation," he said. "Relevant individuals in the United States should respect the facts, discard prejudice, and stop smearing and discrediting China's achievements in the development of its artificial intelligence industry." The U.S. lawmakers said they believe that the threat is real and that AI firms need to be able to share information to counter it. "We cannot allow China, or any foreign adversary, to steal American technological advancements in AI, and then weaponize those advances to undermine our own national security," Schiff, whose state is home to the leading U.S. AI labs, said in a statement. "As new AI innovations are developing faster than ever, it's critical that Republicans and Democrats continue to work together to address potential misuse of AI development by hostile actors at home and abroad," he said. Latta, chair of the House Energy and Commerce subcommittee on energy, said AI models are now critical infrastructure like the electric grid. "Protecting the American people means ensuring the systems they rely on every day, including our hospitals, water systems, and electric grid, remain safe, secure, and reliable," he said in a statement. "Our adversaries are constantly looking for new ways to target our critical infrastructure, and we must make sure we are prepared to respond." Antitrust laws limit coordination within an industry so businesses can't agree to fix prices, divvy up territory or otherwise restrain trade. But Congress has on occasion expressly created exemptions when it has identified national security interests, as it did in 2015 when it passed an exception related to cybersecurity information-sharing. The Justice Department's Antitrust Division also has the authority to grant temporary or limited exemptions for businesses to coordinate, a power it used in 2020 to allow vaccine makers to share information early in the Covid-19 pandemic. But within the tech industry, the value of distillation is hotly disputed, especially as customers face big bills from the leading U.S. AI labs. Arguing in favor of using low-price Chinese models, Nic Carter, a Miami-based venture capitalist, said in a widely shared post on X this week that American businesses and consumers "will benefit from hyperdeflation in the cost of digital cognition just like everyone else." Another venture capitalist, Bill Gurley, said on X that AI firms should sue, rather than lobby the government, if they think their work is being unlawfully infringed upon. Some U.S. companies have been vocal in their defense of using foreign AI models. Hugging Face, which makes software used by AI labs, said last week that it turned to a Chinese model to successfully analyze a large-scale cyberattack after "frontier models" from the U.S. were no help. On Tuesday, OpenAI revealed in a blog post that one of its autonomous agents was behind the hack after it went rogue. Clement Delangue, the CEO of Hugging Face, praised distillation on X in April. "What people call 'distillation' is a super common practice (you use other models to benchmark your model, to evaluate your inputs or to add a little bit to your datasets) that in my opinion should be covered by fair use (just like using public data is), especially when the resulting models are open-source as it benefits all and helps break out monopolies that are strongly forming in AI!" he wrote. SpaceX CEO Elon Musk testified in April as part of his lawsuit against OpenAI that his AI researchers had used distillation techniques on OpenAI models to train his chatbot, Grok. Google offers a limited distillation service so enterprise customers can run their own smaller, cheaper versions of the Gemini chatbot.

OpenAI's Brockman Says Kimi K3 Is 'Pretty Good,' Won't Confirm Distillation

Brockman called Kimi K3 "pretty good" but said it is too early to tell if Moonshot distilled from OpenAI, as the White House accused the startup of IP theft OpenAI president Greg Brockman called Moonshot AI's Kimi K3 "pretty good" in a Bloomberg interview on Tuesday, a rare public acknowledgement from a US frontier lab that a Chinese rival has produced a competitive model. Brockman said it is "too early" to determine whether Moonshot trained K3 by distilling outputs from OpenAI's own systems, a practice the company has repeatedly accused Chinese labs of using. His measured tone stood in sharp contrast to the White House, which hours earlier had levelled its most specific accusations yet against the Chinese startup. OSTP Director Michael Kratsios accused Moonshot of accessing banned Nvidia chips through intermediaries in Thailand and of conducting large-scale distillation against American AI models, according to Bloomberg. The accusations landed the same day that Treasury Secretary Scott Bessent said the administration would scrutinise Chinese open-source models for stolen intellectual property and could impose sanctions within weeks. Nvidia CEO Jensen Huang pushed back against the alarm, calling Chinese open models "excellent" and urging the government to let American firms use them. The distillation charge has been building for months. In February, Anthropic accused Moonshot, DeepSeek, and MiniMax of running roughly 16 million exchanges through about 24,000 fraudulent accounts to extract knowledge from its Claude models. OpenAI has made similar complaints, and the White House now appears to be treating the allegations as a matter of national security rather than a commercial dispute between rival labs. Brockman framed OpenAI's position around the company's infrastructure investments, arguing that the US has a "huge advantage" because of the billions it has poured into compute. He said open-weight models like K3 are not truly free because running them at scale still requires expensive hardware, a point that undercuts the narrative that Chinese labs can simply give away frontier-level AI. He estimated that China remains roughly four months behind the US in overall model development. K3 is a nearly three-trillion-parameter open-weight model that Moonshot released on July 16, and it has rattled Washington and Wall Street by matching or beating the best American systems on several benchmarks. The model topped Arena's frontend coding leaderboard within 24 hours of launch, placing ahead of both Anthropic and OpenAI's flagship offerings. Moonshot is pricing K3 at a fraction of what US labs charge, accelerating the commoditisation of the model layer that investors have been warning about since DeepSeek's price cuts earlier this year. Brockman, who donated to a Trump-aligned super PAC, declined to wade into the debate over whether Chinese models should be banned. He instead stressed that OpenAI believes open models are important for "democratizing access to AI," a position that aligns him more closely with Huang than with the White House officials pressing for sanctions.

Senior White House official accuses Moonshot AI of copying Anthropic's leading frontier model

A senior White House official has accused Chinese artificial intelligence startup Moonshot AI of using so-called "distillation" techniques to essentially pirate the capabilities of Anthropic PBC's most powerful models in its own, open-source AI systems. "We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," said Michael Kratsios (pictured), director of the White House Office of Science and Technology Policy, in a post on X on Wednesday. Kratsios was referring to Moonshot AI's new Kimi K3 model, which was pronounced as the world's largest open-weights model when it was released last week to great fanfare due to its prowess, which matches the capabilities of Anthropic's and OpenAI Group PBC's best AI systems. He added that the company went to great lengths to try and hide what it had done. "To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection," he claimed. Experts say Kimi K3 is the most advanced open-weights model released so far, performing at the same level, or sometimes even above that of other leading frontier models in multiple benchmarks. But whereas Anthropic and OpenAI intend to charge for access to their best models, having spent billions of dollars to develop them, Moonshot AI plans to make Kimi K3 freely available for anyone to download, run and customize on their own servers or devices. What did Moonshot AI allegedly do? Kratsios claims that Moonshot AI used a machine learning compression method known as distillation to transfer knowledge from Fable 5 to Kimi K3. In distillation, a larger "teacher" model is used to train a smaller "student" model. Rather than learn only from basic right-or-wrong answers, the student model learns from the teacher model's full probability distributions - for instance, a 90% chance for word A, 8% for word B - which reveal clues about how the teacher thinks. Distillation is sometimes used in "legitimate" ways, Kratsios said. For instance, some AI labs use the technique to train less powerful models on the outputs of a more powerful one to make more efficient versions of the same AI systems. "Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem," he said. "However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable." Kratsios's comments come just days after U.S. Treasury Secretary Scott Bessent said that the U.S. government could sanction Chinese AI model makers if proof emerges that they basically copied American-made models to develop their own. They're notable, because they represent a rare instance in which the Trump administration has appeared to be supportive of Anthropic. There has been a lot of antagonism between the White House and the AI firm recently, with officials recently preventing the company from releasing Fable 5 due to security concerns. Prior to that, in February, the Pentagon blacklisted Anthropic after it was refused full access to its leading frontier models. "Thanks @mkratsios47 for speaking out on this important issue," wrote Sarah Heck, Anthropic's head of public policy, in a reply to Kratsios's post on X. Heck added that Anthropic is continuing to work with the government to try and prevent U.S. models from being distilled by Chinese firms. "Illicit, adversarial distillation is IP theft and industrial espionage that supports adversary military and intelligence capabilities," she wrote. "It is a national challenge that creates serious national security risks for the United States and democratic allies." Anthropic first accused Chinese firms of abusing the distillation process in February, pointing fingers at Moonshot AI, MiniMax and DeepSeek Ltd. and calling on the U.S. government to take action to prevent it. But while White House officials paid lip service, Anthropic did not get much sympathy from the general public, which noted the irony of the allegations, given that it trained its systems on massive volumes of copyrighted data posted online. Experts are skeptical Despite getting Kratsios's support, there are many who believe that distillation cannot be responsible for the advanced capabilities of Kimi K3. For instance, Braden Hancock, a researcher at the Laude Institute and the co-founder of AI startup Snorkel AI Inc., told TechCrunch that it would be almost impossible to train a model as strong as Kimi K3 on Fable 5 so quickly. "There's just not even frankly time, right?," he said. "Fable's only been publicly available since July 1st. You can't distill that much data, train a model, and release it in two weeks." Another expert, Nathan Lambert, who is a researcher at the Allen Institute for AI, was similarly skeptical that Moonshot AI could have pulled it off so quickly. "I've been of the opinion that distillation has becoming less and less impactful over time as the Chinese models get closer to the frontier and the training regime shifts to [reinforcement learning]," he said in a podcast this week. "[If] it were the case, everyone would be easily able to catch up to a GLM or to a K3 by using its data for distillation. But we have not, or we won't see this, from supervised fine-tuning alone." Given that the White House claims to have "evidence" of Moonshot AI's guilt, the next question is whether or not the government will actually follow through on Bessent's threat to start sanctioning Chinese AI companies. Moonshot AI has not commented on the allegations, but a spokesperson at the Chinese embassy in Washington DC told Reuters that the claims are "entirely unfounded." "Relevant individuals in the United States should respect the facts, discard prejudice and stop smearing and discrediting China's achievements in the development of its artificial intelligence industry," the spokesperson said.

US Accuses Moonshot AI of Covert Anthropic Model Distillation

US officials warn that Chinese firms conducting covert, industrial-scale AI distillation could face sanctions and export restrictions. A White House official accused Moonshot AI of distilling Anthropic's Fable AI model to develop Kimi K3, which launched last week. In a post on X on Wednesday, White House Office of Science and Technology Policy Director Michael Kratsios alleged the Chinese AI firm developed an internal platform to distill US models at scale, using methods designed to evade detection. "Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem," he said. "However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable." Kimi K3 has emerged as one of China's most capable AI models, intensifying Washington's concerns that American models are being covertly used to accelerate China's AI progress. However, some AI researchers questioned claims that Anthropic's latest AI model was used to train Kimi K3. Anthropic's Fable 5 was re-released on July 1 after it was quickly taken offline due to US export controls, while Kimi K3 launched on July 16, giving a narrow window for distillation attacks to occur. "There are only 15 days between fable 5 ban removal and kimi K3 release," said Elie Bakouch, a researcher at AI startup Prime Intellect. "I don't think claiming that K3's performance comes from fable distillation (even if they did it) makes sense technically." Dean Ball, OpenAI's head of strategic futures, said on Friday he didn't believe the K3 model's performance could be "explained away by distillation or anything like that." "We support open-source AI and the innovation it unlocks. But open source is not open season on American IP," said US Treasury Secretary Scott Bessent. "When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table."

White House accuses Moonshot of copying Anthropic AI

White House science advisor Michael Kratsios alleged that Moonshot, a Chinese firm, copied Anthropic's Fable large language model (LLM) to create the Kimi K3, which is the largest open-weight model currently available. Kratsios stated Moonshot's model utilized chips not authorized for export to China, adding that the covert industrial distillation aimed at stealing U.S. technology is "unacceptable." He noted ongoing discussions about the potential banning of Chinese open-weight models that have become controversial in the AI sector. Treasury Secretary Scott Bessent echoed Kratsios' comments, revealing that U.S. large language models have been detected in many Chinese models, which he also described as "unacceptable." The Treasury Department did not clarify the exact nature of the watermarks found in these models. Experts, however, expressed skepticism about whether distillation alone could account for the advanced capabilities of Kimi K3, especially considering Fable's release date of July 1. Braden Hancock, a researcher at the Laude Institute, pointed out that it is implausible to develop a model with the strength of Kimi K3 in such a short time frame using only distillation techniques. He stated, "You can't distill that much data, train a model, and release it in two weeks." Nathan Lambert, from the Allen Institute for AI, noted the diminishing effectiveness of distillation over time and argued that advanced methods like reinforcement learning might be necessary for such models to achieve their capabilities. The distillation process involves systematically querying an LLM to replicate its functionalities. This process might include supervised fine-tuning (SFT), where the target model trains a new model by emulating its behaviors. Lambert assessed that while SFT has benefits, they are waning as model complexity rises. To replicate Fable's capabilities might require reinforcement learning, which can be resource-intensive and may not guarantee superior performance. An ongoing conflict exists in the AI industry regarding distillation practices. Anthropic had earlier accused Moonshot of systematically extracting data from its models, citing millions of suspicious exchanges monitored through IP addresses. Elon Musk testified that his company, SpaceXAI, also utilized distillation techniques on OpenAI models to develop Grok, demonstrating how common this practice is across the sector. Hancock underscored that American perceptions may downplay the technical skills of Chinese AI teams, referring to Moonshot's co-founder, a PhD graduate from Carnegie Mellon University. Moonshot allegedly leveraged banned Nvidia chips while accessing GB300-equipped servers in Thailand to develop Kimi K3. Sam Bresnick, a research fellow at Georgetown's Center for Security and Emerging Technology, noted the existence of a black market for such chips, which remain banned for export to China. Efforts to enforce regulations on the use of these advanced chips have yet to see significant progress. President Biden's Department of Commerce proposed rules for data centers to establish customer identification measures in 2024. However, no significant developments in this area occurred during the previous administration, and exporters must ensure their advanced chips are used for legitimate applications.

China says AI development comes from 'greater self-reliance and strength' amid stolen tech claims

China defended the country's development of artificial intelligence, stating it comes from "greater self-reliance and strength" after a top White House official accused a Chinese startup of improperly using Anthropic's latest model and accessing restricted Nvidia chips. Lin Jian, China's Foreign Ministry spokesperson, told reporters on Wednesday the country "opposes politicizing and instrumentalizing trade and tech issues," as the startup Moonshot AI faces allegations of distilling technology from U.S. frontier labs to develop its latest model, Kimi K3. "The development of AI in China comes from greater self-reliance and strength in science and technology, and is fueled by China's vision of extensive consultation and joint contribution for shared benefit," Lin said, adding later the politicization "will only stifle global AI advances and serve one." In a separate statement to NBC News, Liu Chang, a spokesperson or the Chinese embassy in Washington called the allegations unfounded "smear." "The so-called allegation about China 'stealing' intellectual property developed by U.S.-based AI and life science startups is entirely unfounded," he reportedly said. "China has consistently attached great importance to the protection of intellectual property and is committed to safeguarding, in a fair and impartial manner, the lawful rights and interests of both domestic and foreign IP holders." Kimi K3, released late last week by the China-based firm Moonshot AI, quickly sent shockwaves through industry and policy circles, threatening the leads in coding ability of U.S. labs and models like Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol. Michael Kratsios, the director of the White House Office of Science and Technology Policy, claimed on Wednesday the administration has information that Moonshot AI distilled Anthropic's Fable to produce Kimi K3. Distillation occurs when the outputs of a stronger model are used to train less capable models. "To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection," Kratsios wrote on social platform X. The White House official also accused Moonshot of obtaining servers equipped with Nvidia's GB300 system and accessing the same systems in Thailand, and suggested they were likely used to train the firm's AI models. "The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models," Kratsios said. "Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem," he continued. "However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable." A day earlier, Treasury Secretary Scott Bessent warned the Trump administration could sanction China over IP theft. "We've seen a lot of talk about ... open-source models coming and threatening the large language models in the U.S.," Bessent told Fox Business's Maria Bartiromo. "What I will tell you is this administration supports open-source models, but what we do not support is IP theft," he continued. "And if we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft." The Trump administration previously called out China in April, when Kratsios said the U.S. government is aware Chinese entities allegedly leveraged thousands of proxy accounts and jailbreaking tactics to access proprietary information. It is not clear what evidence the government regarding the creation of Kimi K3, but multiple tech and policy analysts voiced the same suspicions to The Hill earlier this week. Hamish Low, a research associate in compute policy at the Institute for AI Policy and Strategy, reviewed Kimi 3 and told The Hill that it's "pretty overwhelming likely they are distilling from leading U.S. models." He pointed to how Kimi is excelling in coding but not necessarily in cybersecurity. This is "pretty much what we would expect if they were doing lots of distillation of coding data, which they can more easily access from these leading models," Low said, adding Kimi is not on par with U.S. companies, which are more heavily "safeguarded" by the U.S. firms.

U.S. Treasury Secretary Bessent threatens sanctions against Chinese AI model makers

U.S. Treasury Secretary Scott Bessent said today that the White House is going to examine some of the highest profile "open-weight" models from China to see if their creators have stolen the intellectual property of America's top artificial intelligence frontier labs. If evidence suggests that's the case, the U.S. could respond with sanctions against the Chinese firms found guilty, he said. In an interview with Fox Business, Bessent (pictured) said the Trump administration supports open-source models, but is dead against IP theft. "If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," he said. Bessent's comments come just days after one of China's hottest AI startups, Moonshot AI, unveiled a new open-weights model called Kimi K3 that matches the performance of OpenAI Group PBC's and Anthropic PBC's most advanced models. Chinese models are becoming increasingly popular with U.S. businesses, because they're seen as more than good enough in terms of their capabilities, and cost far less money thanks to their open-source nature. It's a trend that threatens to derail the business models of America's top AI labs and prevent them from raising further capital. On Monday, a report in Axios claimed that the Trump administration was even considering implementing a wholesale ban on open-source models from China, though officials have since distanced themselves from that report. Both Anthropic and OpenAI have publicly voiced concerns in recent months that foreign competitors are harvesting massive amounts of data from their models in order to use it to build their own, redeployed as open source technology. In April, the White House vowed to work with them to try and combat this perceived theft. If the U.S. does implement sanctions against Chinese model makers, it would be a significant escalation of an increasingly bitter race to maintain leadership in the AI industry. The U.S. has already taken substantial steps to rein in China's capabilities, restricting its access to the most advanced AI processors. But the sanctions would mark the first time it has targeted Chinese model makers directly. China's leading AI firms are accused of using a technique known as "model distillation," which makes it possible to assimilate the capabilities of much larger models into smaller systems that are easier to run. The technique has been used by many American companies to create more efficient versions of their own models, and not everyone agrees it constitutes theft. One of the biggest critics of those criticizing China's AI labs is Microsoft Chief Executive Satya Nadella, who insinuated that it's hypocritical of American firms to make such accusations. "While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation," he wrote in an article on X last month. It's hard not to see Anthropic just as guilty as the Chinese firms it's accusing, and U.S. law appears to agree that's the case. Earlier this week, the Claude chatbot creator revealed it would start writing checks to authors as part of a $1.5 billion settlement it had agreed after a judge ruled that it had illegally downloaded and stored millions of copyrighted books that were used to train its models. Meanwhile, OpenAI is still embroiled in its own, long-running feud, having been sued by the The New York Times in 2023 for allegedly training its models on its content without permission. Any sanctions imposed by the U.S. on Chinese AI firms may well be moot, however, if Beijing decides to make good on a recent threat of its own. Earlier this month, Reuters reported that Chinese officials are considering restricting the ability of foreigners to access its most advanced models, in response to similar moves made by the White House.

As AI grows more powerful, a US-China feud threatens safety efforts

US warnings of sanctions targeting Chinese AI companies could jeopardize critical safety conversations. The rise in allegations of intellectual property theft and breaches of export controls contributes to rising tensions. With advanced AI technologies posing serious security risks, both nations recognize that frontier AI is a double-edged sword. Urgent cooperation on establishing safety protocols is pivotal for fostering responsible advancements in AI technology. BEIJING: U.S. threats to sanction Chinese AI developers over alleged intellectual property theft and export-control violations risk unravelling efforts to establish a bilateral dialogue on AI safety, analysts say, just as increasingly powerful models raise concerns over serious security breaches. U.S. officials on Wednesday accused Chinese AI lab Moonshot of distilling its Kimi K3 model from Anthropic's advanced Fable 5 model, prompting Treasury Secretary Scott Bessent to warn of potential sanctions. Distillation is the process of training AI models using output from more advanced ones to lower the costs of training a powerful new AI tool. The Commerce Department's Bureau of Industry and Security is also investigating whether Chinese firms like Moonshot are accessing advanced U.S. chips illegally to train their models. A Moonshot spokesperson did not respond to Reuters' request for comment. The dispute exposes a growing tension at the ⁠heart of ⁠the U.S.-China AI race. Both sides increasingly view frontier AI as a strategic asset and a security risk, yet experts argue cooperation on safety standards is becoming more urgent as powerful models could be misused for cyberattacks or other harmful applications. The accusations, coupled with reports Washington is weighing restrictions on Chinese open-weight models, could provoke tit-for-tat measures and jeopardise a planned U.S.-China AI dialogue in September, analysts say. The tensions come atop years of U.S. export controls designed to curb China's access to cutting-edge chips, measures that have increasingly dominated trade talks between the two countries. "Depending on the number of Chinese companies targeted, (and) the nature of the punitive actions taken... the retaliation has the potential to scuttle both the AI dialogue and the September 24 meeting between Presidents Trump and Xi," said Paul Triolo, a partner at DGA-Albright Stonebridge Group. Beijing meanwhile is considering whether to restrict overseas users' access to its models as a ⁠retaliatory measure, Reuters previously reported. CLOSED AND OPEN-SOURCE SECURITY RISKSNew York-based Hugging Face last week used a Chinese model, Z.ai's GLM-5.2, to contain a cyberattack by a rogue OpenAI agent that escaped during safety testing, since American closed models' safety guardrails were ironically too strong. As Chinese and U.S. frontier models achieve recursive self-improvement (RSI), where systems can ⁠autonomously enhance their own capabilities, researchers say both countries have an incentive to cooperate on safety standards before a more serious incident occurs. Industry experts are also highlighting the risks posed by open-weight models, which can be downloaded, modified and redistributed with little oversight, limiting the effectiveness of software-based export controls. "Godfather of AI" Yoshua Bengio warned at China's flagship AI forum last week that the deployment and sharing decisions of open-weight models are irreversible, and their safeguards are easier to remove. "The logical thing to do is to find a good evaluation of these models, share the models that are not too dangerous, and not share those above the threshold of risk," he said via videolink. RESOURCE, GEOPOLITICAL CONSTRAINTS Chinese AI models must undergo government safety and content reviews before release, though current rules do not cover post-release modifications. Unlike U.S. giants OpenAI and Anthropic, many Chinese AI labs lack the computing resources for extensive safety training and instead focus on improving model capabilities, industry observers say. "It's plausible China reconsiders allowing open-weight releases for frontier models down the line, but the bar for doing so is high," said Kristy Loke, a MATS research fellow studying China's AI governance. Some AI safety researchers argue that frontier models in both the U.S. and China should face stricter pre-release testing and independent third-party evaluations. In the U.S., firms can voluntarily submit models for testing ⁠by the state-funded Center for AI Standards and Innovation (CAISI), though some lawmakers have called for mandatory federal reviews. "In an ideal world, the two countries will come together to work on safer models... agree to build common standards around pre-release testing and set red lines for the most advanced open models," said Loke. DIVIDED RESPONSE ON CHINESE AI MODELSSome U.S. government and AI industry officials remain split on how to respond to Chinese models, which account for about 60% of token usage by U.S. companies on the OpenRouter platform. OpenAI and Anthropic have lobbied Washington against lower-cost Chinese models, arguing they could undermine their business models. OpenAI lead strategist Dean Ball wrote on X that the Trump administration could "create large amounts of regulatory risk around the use of open-weight Chinese models" to curb their adoption by U.S. firms. White House AI adviser David Sacks countered that leading U.S. labs "want the government to eliminate their open-source competition" and argued for fair competition, adding in another post that the "Kimi Panic needs to stop". "As long as we don't sabotage ourselves with unnecessary rules, the U.S. will continue to win," he said.

White House official accuses Chinese startup of distilling Anthropic model, accessing banned Nvidia chips

A top White House official accused a Chinese startup on Wednesday of improperly using Anthropic's latest model and accessing restricted Nvidia chips to develop its new Kimi model. Michael Kratsios, the director of the White House Office of Science and Technology Policy, said the administration has information that Moonshot AI distilled Anthropic's Fable to produce Kimi K3. Distillation refers to a technique used to transfer the knowledge of a larger, more capable model to a smaller model. "To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection," he wrote in a post on X. Kratsios also alleged that Moonshot has acquired servers equipped with Nvidia's GB300 system and has accessed the same system in Thailand, suggesting they were likely used to train the firm's AI models. The GB300 is some of Nvidia's most advanced technology, containing its Blackwell chips, which are not allowed to be sold to China. The highest-level chip currently receiving licenses from the administration is the H200. "The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models," Kratsios said. "Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem," he continued. "However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable." The release of Moonshot's Kimi model last week, which the company claims can perform on par with top models from Anthropic and OpenAI, has raised concerns about whether China is closing in on America's lead in AI. Treasury Secretary Scott Bessent separately warned Tuesday that the Trump administration could sanction China over intellectual property (IP) theft if they are found to be "stealing from our great companies." "We are finding watermarks of our U.S. large language models on many of the Chinese models, and that that's unacceptable, Maria," Bessent told Fox Business's Maria Bartiromo. "So, we're going to be looking at that in the coming days or weeks."

White House official says Moonshot accessed banned Nvidia chips

A White House official accused China's Moonshot of improperly using U.S. AI models and Nvidia chips to create the Kimi K3 system that stunned the tech industry last week with its advanced capabilities. The company "acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models," White House Office of Science and Technology Policy Director Michael Kratsios said in a social media post Wednesday, making the case that Moonshot violated U.S. export control rules and companies' terms of service to build its product. The semiconductor servers Kratsios referenced are part of the Blackwell generation of Nvidia products, which the U.S. forbids from sale to Chinese companies. Though the Trump administration has allowed some sales of Nvidia's H200 chips, it continues to prevent sales of the more advanced Blackwells.

Scott Bessent Says Open Source AI Not 'Open Season' on American Intellectual Property, Warns Chinese AI F

Treasury Secretary Scott Bessent on Wednesday warned that Chinese artificial intelligence companies could face sanctions and Entity List designations if they use "industrial-scale distillation attacks" that cross into intellectual property theft. Bessent Escalates Warning In a post on X, Bessent said President Donald Trump's administration supports open-source AI and the innovation it unlocks but added that "open source is not open season on American IP." "When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table," Bessent added. Meanwhile, Rep. Ted Lieu (D-Calif.) questioned the administration's decision to threaten sanctions over alleged AI IP theft while allowing the sale of high-performance AI chips to China. US Alleges Moonshot AI Used Anthropic Models Anthropic's Allegations Against Alibaba The company alleged the effort relied on nearly 25,000 fraudulent accounts that generated more than 28.8 million interactions with Claude to help improve Chinese AI systems. Benzinga edge rankings indicate NVDA has a Momentum score in the 71st percentile and a Growth score in the 98th percentile. Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Image via Shutterstock/ Maxim Elramsisy Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

Trump administration may ban Chinese AI models, Axios reports

The Trump administration is exploring ways to limit access to advanced Chinese AI models. This move could benefit American companies like OpenAI and Anthropic. Previous efforts to ban foreign open-source models have been revived due to Kimi's emergence. Cybersecurity concerns and national security advocates are now pushing for these restrictions. A softer approach involving procurement rules and public pressure is also being considered. China calls US visa regulations 'discriminatory', threatens countermeasures The Trump administration is showing signs it could restrict access to advanced Chinese AI models, a move that could cement the lead held by OpenAI and Anthropic, Axios reported, citing people familiar with the matter. Parts of the administration have previously tried to impose de facto bans on foreign open-source models. The recent emergence of Chinese model Kimi has revived those efforts, the report said. U.S. companies have increasingly turned to open-source Chinese models because they are cheaper and, with Kimi's rise, nearly on par with American alternatives. Also read: China calls US visa regulations 'discriminatory', threatens countermeasures What was considered earlier The Commerce Department last year weighed adding several Chinese AI labs to its Entity List, a step that would cut off U.S. access without a license, a source close to the administration said, according to the report. The National Security Agency and the White House Office of the National Cyber Director had also discussed issuing an advisory on Chinese AI lab threats, effectively discouraging U.S. firms from using the technology, the source added. Separately, the White House considered an executive order requiring U.S. companies to guarantee security and accept liability for breaches before hosting Chinese models. Commerce also circulated draft rules last summer aimed at Chinese open-source models under its domestic supply chain authorities, a second source close to the administration said. All these efforts were shelved by administration officials wary of regulation stifling innovation. Sriram Krishnan, a former White House adviser who had opposed such restrictions, has since left his post. With that shift and growing cybersecurity concerns, national security hawks have grown louder, pushing ban momentum higher. Neither the White House nor the Commerce Department responded to requests for comment. A softer approach An outright ban may not be necessary to push U.S. firms away from Chinese platforms. "What's actually happening is slower and more durable," a source familiar with government discussions said, pointing to procurement rules, Entity List threats and public pressure campaigns. Another source described a strategy of highlighting potential backdoors and security gaps in Chinese models rather than pursuing a formal ban. This approach also aims to encourage a more competitive U.S. open-source ecosystem, the source said. Also read: China's Kimi K3 rattles US AI industry However, U.S. open-source alternatives currently lag behind cheaper, more capable Chinese models. Industry pushback White House AI adviser David Sacks weighed in on X on Sunday, warning that "the leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open-source competition." Sacks has repeatedly cautioned against regulatory capture that could favour large U.S. labs at the expense of competition. Leading AI labs or their allies have approached the administration roughly every three to five months with proposals to restrict open-source models, Axios reported. Anthropic and OpenAI have long backed safety regimes that include licensing frameworks, though those calls have largely stayed abstract until now.

US Threatens Sanctions on China Over Alleged AI Model Theft | PYMNTS.com

The President Donald Trump administration will examine whether Chinese AI models have been distilled from their U.S. counterparts, Bessent said. "If we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," Bessent said, adding that the technical term for this theft is known as "distillation," a practice Chinese AI firms have been accused of using before. For example, Anthropic last month accused operators connected to Alibaba and its AI lab of conducting a massive distillation campaign against its Claude models, generating more than 28.8 million interactions with Claude through roughly 25,000 fraudulent accounts between April 22 and June 5. In February, the startup made similar allegations against three other Chinese AI labs, including DeepSeek, Moonshot AI and MiniMax, although on a smaller scale. "The way distillation works is simple," PYMNTS reported June 26." A campaign sends large volumes of carefully designed prompts to a target model and captures its responses. Those responses become training data. The competing model learns to reason and respond in ways that replicate the original, without paying for the research behind it. It is less like hacking a system and more like sitting next to the best student in class and copying every answer they write, at industrial scale." Detection is difficult as a distillation query looks identical to a legitimate one. A developer seeking help from Claude in debugging a function and a campaign systematically extracting Claude's coding behavior would submit the same type of request. The only indicator is pattern, like massive volume, repetitive structures and prompts focused on the same narrow capabilities, coming from hundreds of coordinated accounts in sequence. As organizations integrate AI models into core operations, "the proprietary logic and specialized training of these models have emerged as high-value targets," Google Threat Intelligence Group warned in a blog post earlier this year. Also Tuesday, U.S. Trade Representative (USTR) Jamieson Greer told CNBC the White House is exploring whether China is using unfair practices to gain an advantage in the AI race. "We're taking a very close look at how China is propagating its AI development to make sure that our companies compete again on a level playing field," Greer said. For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

Bessent warns China could face sanctions over AI IP theft

Treasury Secretary Scott Bessent warned Tuesday the Trump administration could sanction China over intellectual property (IP) theft amid concerns about whether Beijing is relying on American AI to develop its own advanced models. "We've seen a lot of talk about you know open-source models coming and threatening the large language models in the U.S." Bessent told Fox Business's Maria Bartiromo. "What I will tell you is this administration supports open-source models, but what we do not support is IP theft," he continued. "And if we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft." The release of Kimi K3, a new model from the Chinese startup Moonshot AI, last week has spurred fears in both Silicon Valley and Washington about America's ability to maintain its lead in the AI race. The company claimed the model could perform on par with Anthropic and OpenAI's leading models. This was reinforced by an AI evaluator's decision to bump Kimi to the top of its rankings for front-end coding, above models from the two U.S. AI labs. Unlike its American counterparts, Kimi is also an open-weight model. While these types of models are not entirely public like their open-source counterparts, the way they are trained to sift through information and formulate answers is made public. The panic over Kimi has once again prompted discussions about distillation, a technique used to transfer the knowledge of a larger, more capable model to a smaller model. "We are finding watermarks of our U.S. large language models on many of the Chinese models, and that that's unacceptable, Maria," Bessent said Tuesday. "So, we're going to be looking at that in the coming days or weeks." "The other question might be: Should companies that are using Chinese models have to disclose that to their customers?" he added.

Scott Bessent Warns of Possible Sanctions on Chinese AI Over Alleged IP Theft: 'This Administration Suppo

Treasury Secretary Scott Bessent has indicated that the Trump administration is set to probe alleged intellectual property theft by Chinese artificial intelligence models. Bessent made it clear on Fox Business' "Mornings with Maria" on Tuesday that the U.S. government supports open-source models but not "IP theft." He cautioned that foreign models found guilty of stealing from American firms could be subjected to sanctions. "What I will tell you is this administration supports open-source models, but what we do not support is IP theft," said the Treasury Secretary. Bessent used the term "distillation" to describe this kind of theft, which refers to a method of AI training where a less competent model is developed using outputs from a more advanced model. He voiced concern over discovering "watermarks of our U.S. large language models on many of the Chinese models," deeming it unacceptable. Bessent confirmed that this issue would be addressed in the near future. As per a Reuters report, the U.S. and China have planned to discuss AI-related issues in September, with Bessent representing the U.S. This development comes amid growing tensions over AI between the U.S. and China. Chinese open-weight AI models are rapidly catching up to leading U.S. systems from OpenAI and Anthropic, raising concerns about America's long-term edge in artificial intelligence. Chinese startup Moonshot AI recently launched its Kimi K3 model, which outperformed several top U.S. models on certain industry benchmarks. Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Image via Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

Scott Bessent sends unprecedented warnings to China on AI models

The U.S. and China have been fighting over AI for a while now. Chip bans, export controls, benchmark competition, open-source model releases that sent shockwaves through Silicon Valley - most of it has been quiet enough that regular people didn't notice. July 22 was different. Scott Bessent got on X (the former Twitter) and said something that caught the attention of every major AI company with international operations. The Treasury Secretary wasn't talking about tariffs or trade. He was talking about how one Chinese AI company trained its model, and what he said was coming if Washington's version of events held up. What Scott Bessent said about Chinese AI companies and American IP Bessent's post had one line that stood out from everything else. "Open source is not open season on American IP," he wrote on X. "When firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table." Entity List. That's the list Huawei is on. That's the list the U.S. puts companies on when it decides they're a national security problem. Bessent just dropped that word into a conversation about how a Chinese AI company trained its model. That's the part of the post worth sitting with. He had made a version of this threat the day before, too, saying the government would examine Chinese open-source models for signs of IP theft. The July 22 post was a doubling down, not a first shot. Why Moonshot and Kimi K3 are at the center of this AI dispute The target is Moonshot, a Chinese AI company that released Kimi K3 on July 16. It's a 2.8 trillion parameter model, the largest open-weight AI model ever built, and Moonshot is promising to publish the full weights by July 27. White House Science and Technology Policy Chief Michael Kratsios pointed the finger at them on July 22, posting on X that Moonshot had conducted large-scale distillation against U.S. models. Kratsios went further than the IP angle. He alleged that Moonshot acquired Nvidia GB300-equipped servers and accessed GB300 systems in Thailand. That's a separate and more serious problem. GB300 servers are part of Nvidia's Blackwell generation and are banned from being sold to Chinese companies. If those allegations are accurate, Moonshot may have violated export-control rules on top of the IP complaint. Moonshot hasn't responded publicly. Some experts are skeptical of the full picture. Fable only became available on July 1, and researchers have questioned whether Kimi K3 could have been built primarily through distillation from a model that's only been out for three weeks, according to TechCrunch. What model distillation actually is and why it's become a legal and political problem Distillation is one of the most common techniques in AI development. A smaller, cheaper model learns from the outputs of a larger, more powerful one. Done properly, it's a legitimate optimization method. Most AI companies do some version of it. The dispute is never really about whether distillation happened. It's about whether it was done at a scale and in a manner that amounts to stealing someone else's work. Most major AI models have terms of service that say you can't use their outputs to train a competing model. If Moonshot was systematically hitting Anthropic's systems to generate training data, that's a terms violation at minimum. Whether it rises to the level of legal IP theft is a much murkier question, and whether Treasury sanctions are the right tool for a terms-of-service dispute is murkier still. The open-source angle makes it more complicated. Moonshot released Kimi K3 as an open-weight model, which has drawn comparisons to DeepSeek's approach earlier this year. That release rattled U.S. AI labs because it suggested Chinese companies might be closing the gap faster than expected. Washington's response appears to be that if the gap is closing because American IP is being stolen, that's not acceptable competition. Why Washington is pushing to restrict Chinese AI models more broadly Bessent and Kratsios aren't the only voices in this conversation. Dean Ball, formerly a White House AI adviser and now OpenAI's Head of Strategic Futures, has been arguing that the U.S. should restrict or effectively ban the use of Chinese open-weight models entirely. His case is partly about national security and partly about preserving the economic model that justifies the enormous capital costs behind frontier AI development. Bessent himself has been framing the U.S.-China AI competition in existential terms for months, saying at a Wall Street Journal event in April that the U.S. had just 12 to 18 months before AI defines daily life, as TheStreet reported. Kimi K3 got attention because it was good, good enough that people started asking whether Chinese labs were closing the gap faster than anyone expected. That's an uncomfortable question for U.S. AI companies that have been spending hundreds of billions on training runs. If a Chinese company can match that output at a fraction of the cost, the whole investment thesis gets harder to defend. The IP theft accusation gives Washington a cleaner story. But the anxiety underneath it isn't really about one model. Anthropic has been lobbying Washington for tougher action against Chinese AI rivals. The company pressed officials earlier this year for stronger IP protections. In June, it sent a letter to the White House and U.S. senators alleging that Alibaba ran 28.8 million exchanges against its Claude models using 25,000 fake accounts in what it called the largest distillation attempt it had documented. The Fable-Kimi K3 situation is, in some ways, the fight Anthropic was already trying to get Washington to take seriously, as TheStreet reported. What Bessent's AI sanctions threat means for the global AI industry If the U.S. begins imposing sanctions over model training practices, it changes the rules of the AI race in a meaningful way. Companies building models with international teams, using open-source tools, or operating in multiple jurisdictions would have to think carefully about what they're training on and where the data is coming from. Entity List designations are particularly significant. Being placed on the Entity List cuts off a company from U.S. technology supply chains. For an AI company that depends on Nvidia chips, cloud infrastructure, or American software tools, that's potentially a company-ending designation. The broader implication is that AI intellectual property is becoming a national security issue. Washington is no longer treating model training as a purely commercial or technical dispute. When Treasury starts using the same language it uses for financial sanctions in an argument about how one AI company trained its chatbot, the AI industry is operating in a different environment than it was a month ago. The Arena Media Brands, LLC THESTREET is a registered trademark of TheStreet, Inc. This story was originally published July 23, 2026 at 2:17 PM.

Why is OpenAI Scared of Open Source? Both Chinese and US AI Models

The question that we are grappling with is this: Is this fear about money and the public offer or genuinely over national security and war games? Ever since Moonshot AI's Kimi K3 appeared on desktops, the debate around open-weight large language models and its impact on the business prospects of US AI startups and the future of LLMs as a technology has heated up considerably. Not surprisingly, it is OpenAI that has sought protection, albeit in a fashion that raises distrust around the new models. Dean W. Ball, OpenAI's head of strategic affairs took to his X handle asking the Donald Trump administration to find a pretext to create regulatory fear, uncertainty and distrust around the new Chinese models. However, he is against a ban but says "You just need to direct every agency to issue soft law that creates FUD (fear, uncertainty and doubt)." He even gives the White House an example of what that narrative could be. "A Federal Reserve Advisory Bulletin found that there may be backdoors in Chinese AI models" that need not be well-justified. "You can create regulatory risk that every regulated enterprise backs off," while being careful not to create so much that the hyper-scalers back off from Chinese models. Of course, the results were as expected. Some media reports said Trump was close to banning the Kimi K3 in the US while others said the Department of Commerce may refrain from taking such a drastic step. As for fellow tech leaders, they took on Ball claiming that open software can only accelerate innovation and coexist with proprietary projects. Ball had claimed that open-weight models were inherently 'decelerationist' and that he was surprised why it was being called accelerationist or how people could get excited about open-weight models. "I suspect the reason they are in that they know open-weight models are effectively ungovernable, and they simply like pushing this cloak" over the entire AI ecosystem. Among the first to squarely criticise the OpenAI leader's claims was David Sacks, an external advisor to President Trump and a tech founder and investor. "The weaponisation of regulatory uncertainty as a competitive tool should be completely unacceptable, he said via his X handle. Sacks also counters Ball's narrative of not banning Chinese models while creating necessary FUD to drive people away. "Regulatory decisions should always be well justified and grounded in facts, logic, and evidence, not the deliberate exploitation of fear and uncertainty. Implementing a surreptitious policy through manufactured doubt -- rather than strong and explicit justification -- corrodes the rule of law and invites future abuse against anyone," he notes. Given the ferocity of the attack around a possible ban on Chinese AI models, OpenAI's Ball did a backtrack by first clarifying that he'd joined the company only two weeks ago. He quoted a post from the past where he'd said that right now AI isn't too dangerous for it to justify a drastic shift in public policy (such as a ban). Which brings us to the moot question around what exactly could be scaring OpenAI and probably Anthropic and to some extent Google too? Open-weight models running on independent AI infrastructure or within enterprises provides cheaper intelligence compared to the proprietary LLMS and when users spend more outside these closed labs, the returns on massive investments on model training would get smaller. In fact, smaller AI companies in the US itself are now looking at creating frontier calibre-open source models, which could shave off the revenues of these big tech AI giants. It is not about actual AI usage actually reducing, but there would be better return-on-investments for the enterprises that have spent big dollars on AI without really getting the results. In other words, the problem statement emanating from the Chinese models or even the smaller models being generated by smaller US startups is limited to the likes of Anthropic and OpenAI whose revenue numbers may not spike as much as the hype leading up to their public offerings. Which is why narratives are being encouraged now. The most important aspect of such a narrative would be concerns over protecting US data from the Chinese government (remember the ban on Chinese EVs?). However, experts are clear that no such leak would occur if Chinese AI models run on US servers within the country. And if there is a fear in this regard, wouldn't it also be valid for Europe and India to fear a similar threat from American AI companies? Then there is talk about a communist AI, which we believe is also a specious argument in case we are to believe the other debate about data distillation going on. How can a Chinese AI model that is distilling data from US companies then give it a communist hue? If so, what exactly would this mean for activities such as coding? Another refrain around US concerns relate to Chinese models lacking guardrails that the White House has mandated through a process that is far from transparent. These are supposed to prevent leading LLMs (from Anthropic, OpenAI) from being used to exploit closed computer systems or generate autonomous weapons. What we are finding out now (an example is the Hugging Face hack) is that these guardrails are actually making US companies more vulnerable. In fact, David Sacks had shared a few cases where US companies turned to Chinese models to close security gaps when they failed to do so using models made in the United States. "Kimi K3 just fixed 15 critical security bugs that Codex and Fable refused because of "cyber guardrails." There's no reason to limit American models on tasks that Chinese models handle without issue. We're only making ourselves less competitive," he said via his X handle. While all these challenges may work for building an anti-China narrative, the real reason could lie elsewhere. What if Chinese models outpace the US models if the frontier labs actually slow down their growth in the near future? And the reasons could be more military than economic, given that the Pentagon has repeatedly requisitioned help from Anthropic in recent battles. Advocates of open-weight models in the US such as Braden Hancock, the co-founder of Snorkel AI, believe that the bigger impact of open source models coming from China is not about them sneaking in through the back door but actually owning the innovation. "You end up with, effectively, an expanded workforce on your model. PyTorch became the industry standard because it was open source, and so the whole community could contribute to it rather than just one company, and it grew and grew, and all the rest of the deep learning libraries kind of died in comparison," he told TechCrunch. Hugging Face CEO Clem Delangue also believe that restricting open models won't make AI safer. "It would simply hide the risks, concentrate power in the hands of a few and make it harder for the next generation of builders, researchers, academia, non-profits and governments to participate in making AI safer and beneficial to all. Moreover, it would be worth noting that the problems facing US companies in AI are pretty much the same in China too. AI companies are struggling to generate revenues and access compute power in both geographies. Which is why, recent statements from Thinking Machines Lab and Nvidia about open models is making sense at this juncture. Because companies like Nvidia would be better off if they had hundreds and thousands of companies building AI rather than two or three, who have large enough purses to create their own AI chips, like Google has done and is all set to enhance its product in the near future. Maybe, the problem is limited only to the AI giants who have built a business with little relevance to its customers. And now they are facing the music from the Chinese - pretty much the same way that the likes of Elon Musk's EV business did when the world opened up. Protectionism can never be a solution to product quality and innovation. We hope OpenAI gets it and so does Anthropic and at a broader level the American policy makers.

Washington doubles down on its Al fight with China

A blockade is aimed at a foreign capital. The bill tends to arrive somewhere closer to home. That is the part of trade policy nobody puts in the press release, and it has been the quiet subtext of four years of American chip rules. Advanced semiconductors made a near-perfect target. They are physical, expensive and manufactured inside a handful of buildings on this planet, which makes them the closest thing the technology industry has to a chokepoint. Washington squeezed it, and U.S. export restrictions have erased Nvidia's (NVDA) direct market share in China, chief executive Jensen Huang has said. Then in June the same machinery reached past the hardware. The Commerce Department ordered Anthropic to cut off its two newest Claude models for every foreign national on earth, including the company's own non-citizen employees, and both went dark worldwide inside 90 minutes, according to Forbes. Software had moved inside the export control perimeter. Now the perimeter is stretching again, and this time it points at code that is free. The Trump administration is weighing a move that could effectively ban cutting-edge Chinese AI models, a step that would hand American companies like Anthropic and OpenAI enormous influence over the market, according to NewsNation. Weiquan Lin / Getty Images Why American companies switched to Chinese AI models The pressure behind that decision never came from Beijing. It came from procurement departments. Chinese open-weight models are cheaper and, since Moonshot AI released Kimi K3 last week, close enough in quality that the trade-off has stopped being obvious, according to Axios. More Artificial Intelligence: Some firms moved a while ago. Lindy chief executive Flo Crivello shifted 100% of his startup's traffic to DeepSeek and watched the cost curve "crash to the ground," he told CNBC. That is the constituency a ban would land on. Not Moonshot, which publishes Kimi K3's full weights on July 27 whatever Washington decides, according to Tom's Hardware. What a Chinese AI ban would actually look like What struck me reading the reporting is how rarely the word ban appears in it. Commerce considered adding multiple Chinese AI labs to its Entity List last year, a step that would cut off U.S. access without a license, a source close to the administration told Axios. The National Security Agency and the White House Office of the National Cyber Director weighed an advisory steering American firms away from Chinese AI labs, and the White House considered an order making U.S. companies liable for breaches if they hosted Chinese models, according to Axios. Officials who did not want regulation slowing domestic innovation killed all of it. Then the staff turned over, the national security hawks got louder, and Kimi K3 landed. "What's actually happening is slower and more durable," one source familiar with the discussions told Axios, describing procurement rules, Entity List threats and pressure campaigns aimed at the American companies doing the buying. How the AI fight with China escalated in eight weeks When I lined the dates up, the compression was the part that stopped me. Eight weeks separates a voluntary framework from a possible blacklist. * June 2: an executive order set up voluntary federal pre-release testing of frontier models and explicitly barred any mandatory licensing regime, according to CNBC. * June 12: Commerce ordered Anthropic to suspend Claude Fable 5 and Mythos 5 for all foreign nationals, and the company pulled both worldwide, according to The Conversation. * June 30: Commerce lifted the controls, and both models were back on July 1, according to CNBC. * July 16: Kimi K3 debuted at No. 1 on the Frontend Code Arena, beating Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol in blind coding tests, according to Axios. * July 20: administration officials revived Entity List proposals aimed at Chinese AI labs, according to Axios. Who gains if Chinese AI models get blocked Here is the part worth sitting with. The two companies best positioned to benefit are Anthropic and OpenAI, according to Axios, and one of them spent most of June with its flagship models switched off by the same government. The leading closed labs "want the government to eliminate their open-source competition," White House AI adviser David Sacks wrote on X. Sacks and former White House adviser Sriram Krishnan both argue the move would hand a monopoly to the two largest U.S. labs, according to Tom's Hardware. The counterargument runs on national security rather than market structure. American children should not live under "the algorithms of the adversaries," President Donald Trump said at an AI summit last year, according to NewsNation. Both things can be true. A model built under Beijing's jurisdiction carries governance questions no price list resolves, and blocking it still moves market share to two American firms rather than to American users. Where the AI trade goes from here Chip investors read the news as a threat to pricing power rather than a compliment to demand. Nvidia finished Friday, July 17, at $202.81 with a market value near $4.9 trillion, according to StockAnalysis.com, and Apple (AAPL) briefly grabbed the title of most valuable company in the world that same morning, TheStreet noted. That reflex may have the layers backward. Kimi K3 could be "net positive for essentially every other company" while compressing the roughly 90% inference margins the leading labs enjoy, said Gavin Baker, chief investment officer at Atreides Management, according to Benzinga. Cheaper models mean more inference, and inference still runs on somebody's silicon. The repricing risk sits one layer up, with the American software firms whose margins now assume access to a cheap Chinese model they may not be allowed to keep using. Two dates decide how much of this reaches a portfolio. Moonshot publishes the K3 weights on July 27. Nvidia reports earnings on Aug. 26. Between them sits a question four years of export controls never had to answer. What exactly does a blockade accomplish against something anyone can already download? The Arena Media Brands, LLC THESTREET is a registered trademark of TheStreet, Inc. This story was originally published July 21, 2026 at 12:13 PM.

Trump Administration Seeks Control Over Who Gets OpenAI and Anthropic's Most Powerful AI Models: Report

The Donald Trump administration is reportedly seeking greater influence over which companies and organizations receive early access to frontier AI models developed by the likes of OpenAI and Anthropic. OpenAI, Anthropic AI Access Faces New Oversight The administration's newly launched cybersecurity initiative, Gold Eagle, could serve as a clearinghouse that would allow the White House to determine which organizations receive early access to advanced AI systems, CNBC reported on Friday, citing people familiar with the matter. People familiar with the matter told the publication that future rollouts could require explicit government approval for participating partners. If implemented, the change would mark a significant shift from the current approach, under which AI companies have largely controlled early access to their most powerful models through their own trusted-partner programs. A White House official denied that the administration approves AI releases from private companies, telling the publication that government engagements and testing are voluntary. "The Administration continues to collaborate with all of America's frontier labs to strengthen the security of this technology without stifling innovation," the official said. The official added that "decisions on timing and scope of releases rest entirely with the companies," pointing to Trump's recent executive order on AI. The White House did not immediately respond to Benzinga's request for comments. How AI Companies Have Handled Early Access Until now, OpenAI and Anthropic have decided which enterprises, government agencies and other organizations receive early access to their newest AI models. Anthropic has made its cybersecurity-focused Mythos model available to a select group of partners through Project Glasswing, while OpenAI operates a similar initiative called Daybreak for its cybersecurity model. The administration previously asked OpenAI to limit access to its GPT-5.6 model to trusted partners. The report also said Anthropic's Claude Mythos 5 and Fable 5 were temporarily blocked last month over national security concerns before access was restored following negotiations. Why The Administration Is Getting Involved The reported push comes as the White House attempts to balance national security concerns surrounding increasingly capable AI systems with maintaining the U.S. lead in artificial intelligence. Trump's June executive order encouraged frontier AI developers to voluntarily provide the government with early access to models for security testing. Meanwhile, competition from China continues to intensify. On Friday, Chinese startup Moonshot AI unveiled its Kimi K3 model, which reportedly performs close to OpenAI's GPT-5.6 and Anthropic's Fable models on several benchmarks. Reacting to the development, former White House AI czar David Sacks wrote, "This is how you lose the AI race. The rest of the world won't play by our rules if we bog ourselves down." Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Image via Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

Chinese AI firms that rip off US models could face sanctions, Treasury Secretary Scott Bessent warns

The Trump administration could sanction Chinese artificial intelligence firms if it uncovers evidence that their open-source AI models were built by ripping off US labs, Treasury Secretary Scott Bessent warned Tuesday. "This administration supports open-source models, but what we do not support is IP theft," he said during an appearance on Fox Business. "If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft." As The Post has reported, AI giants like Anthropic, OpenAI and Google have each raised alarms about China-based labs using unauthorized distillation - a technique in which a more advanced model is used to train a new one without permission. Those concerns have accelerated as powerful open-source AI models such as Moonshot's Kimi K3 gain popularity in the US while charging a fraction of the cost for those offered by Anthropic and OpenAI. Bessent said that US officials have found the "watermarks" of US-made AI models within Chinese models - a trend that he called "unacceptable." "We're going to be looking at that in the coming days or weeks," he said. Anthropic directly accused Moonshot of using distillation in February, though it has yet to say if there is any evidence that Kimi K3 was distilled. The Treasury secretary, who will be leading high-level US-China talks on artificial intelligence in September, also raised the possibility that US firms that make use of Chinese models could be required to make public disclosures. Bessent's remarks came as top US officials debate the best way to respond to China's rapid advancements in AI. The conventional wisdom that US labs have a six-to-12-month lead on China has come into question given the recent success of Moonshot and another firm called Z.ai, which released an open-source model called GLM-5.2 earlier this month. Earlier this week, Axios reported that the Trump administration was examining potential ways to crack down on Chinese models, such as potentially adding Chinese labs to the Commerce Department's "entity list" - meaning they could not operate in the US without a license. Dean Ball, OpenAI's head of strategic futures and a former Trump AI official, ruffled some feathers on social media by declaring that the success of open-source AI models could result in what he called "full AI communism, which is precisely what China proposes." Critics of such assertions warn that undue regulation could stifle fair competition in the AI world. White House AI adviser David Sacks fired back at OpenAI's Ball in a lengthy X post, suggesting that he was confessing to a "regulatory capture strategy" meant to shape laws to his company's benefit. Sacks added that the US was "at a critical inflection point in AI policy." "The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open source competition," he wrote. "They have laid their cards on the table. It is time for the rest of Silicon Valley -- the vast majority that still values open competition -- to do the same."

Are Chinese AI Models Threatening Trump's Desire for US Hegemony Over AI?

For President Donald Trump, the battle in Iran could end up becoming the easier of the battles he is fighting on the world stage. Reports of the White House's protectionist measures to save AI companies in the United States comes precisely at the time when Chinese startup Moonshot AI began preparations for an IPO in Hong Kong based on its new open-source AI model's success. In fact, the Chinese startup reportedly paused new subscriptions on a temporary basis as demand for its newly launched Kimi K3 model strained capacity as it continued to upend the Silicon Valley startups in the space. Reports of the Trump administration considering a ban on Chinese open-source AI surfaced over the weekend with a report published by The Information claiming that the idea was discussed as far back as last year. These reports gained credence when Axios claimed yesterday that the Kimi K3 model had stoked conversations about banning foreign open source AI. The reasons aren't tough to fathom. Brian Armstrong of Coinbase took to his X handle to suggest that more and more US companies were moving towards Chinese models in order to curtail their AI spending. "How to keep AI spend flat while token usage grows exponentially: Not with friction and spend alerts. With better defaults, routing, and caching," he said in the post. And how did things get so convoluted for the Trump administration and its MAGA theme? That's the thing right there. OpenRouter data quoted by CNBC earlier this month said Chinese models accounted for 30% of the share of tokens used by US companies each week since February 8. In fact, it rose to as high as 46% during a brief period. Over the past 12 months, this number stood at 11% and fell to 4.5% in the first half of 2025. Former White House AI expert David Sacks claimed on an X post that some of the Chinese models could actually do things that the top US models cannot, largely due to guardrails imposed on the US technology by the White House. "Hugging Face tried using American frontier models to analyze an AI-powered cyber-attack. But the guardrails blocked requests containing real exploit payloads so they switched to GLM 5.2 running locally.." he said. In fact, in an X post a day earlier, Sachs said, ""We are at a critical inflection point in AI policy. The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open-source competition." "We are at a critical inflection point in AI policy. The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open source competition. They have laid their cards on the table. It is time for the rest of Silicon Valley -- the vast majority that still values open competition -- to do the same," he noted in response to another post by Dean Ball, Head of Strategic Futures at OpenAI. The warnings didn't cease there. Last night, investors Chamath Palihapitiya and Bill Gurley warned of severe damage to US companies in case Trump went ahead with the idea of banning Chinese open-source AI. In an X post, Palihapitiya said this is forcing US companies to spend "50-100X more than their competitors abroad" on American tech will make for "financially impaired" American companies, which will eventually lead to problems for the AI labs. Meanwhile, on the other end of the spectrum, Moonshot AI has already engaged financial advisers including Goldman Sachs and China International Capital Corp to discuss their IPO plans, though the timetable remains fluid. Not surprising, given that US AI startups OpenAI and Anthropic too have dilly-dallying with their public issue. The company founded Yang Zhilin in 2023 after pursuing doctoral studies at the Carnegie Mellon University, has emerged as the most-watched Chinese AI firm, raising over $2 billion in May from a clutch of Chinese companies. Total fundraise now stands at around $5.5 billion and the company is now seeking $2 billion in fresh capital on a valuation of $30 billion in June. Moonshot had claimed at the Kimi K3 release that it had performed comparably with leading US models in some technical tasks while independent evaluations too praised its strengths, like Armstrong did in his post. What's interesting is that the timing of the launch came as other Chinese AI firms like Z.ai and MiniMax released their low-cost models to challenge the US firms. Alibaba, which has invested in Moonshot, claimed on Sunday that its Qwen3.8-Max-Preview with 2.4 trillion parameters, had debuted on its AI platforms ahead of a planned open-weight release. Of course, all of this is where the US export controls kick in, given that it has limited exports of advanced Nvidia chips to choke compute power for these Chinese companies. Moonshot's decision to temporarily pause new subscriptions were an obvious outcome of it facing compute challenges. The company claimed that over the past 48 hours, user requests had sharply exceeded forecasts and were approaching the limits on existing clusters. The result is that the company is allocating available compute to current paid users. The company posted on X that it would be splitting future membership into two plans, one of which would be just for coding, a smart move given that most of its users are developers. "Kimi K3 has received far more love than we expected, and our GPUs are feeling it," Moonshot said adding that new subscription spots ⁠would reopen in batches as more capacity was built. Not surprising why these Chinese models are gaining traction among US companies and as they further narrow the gap with leading US rivals, the cost alone will matter. And this is precisely the reason for President Trump to get worried as open source and open weight models could potentially take away the sheen achieved by Anthropic and OpenAI's latest models, in spite of the export bans that the White House initiated before calling it off. These Chinese models make different parts of an AI model available for developers to inspect, use and even modify. Unlike those from closed systems operated by flagship models of OpenAI, Anthropic and Google where the code and the inner workings are proprietary. US enterprise clients find the Chinese model cheaper and more customisable. It remains to be seen whether President Trump sees the larger good of America Inc. to not chase the "ban route" or the "tariff journey" while deciding on the future of AI. Given his penchant for tariffs, we won't be surprised if he does actually push the buttons to restrict compute power. We will keep watching this space as the US-China AI war hots up further.