US-China AI tensions reach new heights as both nations move to restrict each other's models

EXCLUSIVE: Beijing is looking at curbing overseas access to China's top AI models, sources say

SINGAPORE, July 7 (Reuters) - Chinese authorities have held meetings with top tech firms over the past month about potentially restricting overseas access to China's most advanced AI models, including those yet to be released, three people familiar with the discussions said. The talks follow a number of steps by Beijing to keep homegrown AI within the country and underscore how China, like the U.S., is now treating cutting-edge artificial intelligence as a critical national asset that needs controls. Companies present at the talks included tech giants Alibaba (9988.HK), opens new tab and ByteDance as well as startup Z.ai (2513.HK), opens new tab, said the people, who were not authorised to speak to media and declined to be identified. Since the emergence of DeepSeek's R1 model last year, Chinese AI models have made big inroads globally thanks to their low costs and increasing capabilities. Any decision by Beijing to limit access to those products could ripple across AI markets as costs for many businesses would likely increase. TOUGHENING UP PENALTIES FOR AI THEFT DISCUSSED At the meetings, led by China's Ministry of Commerce, participants discussed putting limits on the most advanced AI models -- ⁠both closed-source and more open versions, according to two of the sources. Officials talked about making any leak or theft of proprietary AI technology an offence under China's stringent national security law, one of the sources said. The officials also raised the possibility of implementing new measures to restrict who can fund domestic AI startups, the source added. The scope of the potential restrictions is still being discussed, two sources said, adding that they may only apply to future models. It was not immediately clear when or even if they would come into force. China's commerce ministry, which oversees export regulations, and the National Development and Reform Commission -- the country's state planning agency whose officials also attended the meetings -- did not respond to Reuters requests for comment. Alibaba, ByteDance and Z.ai also did not respond to Reuters queries. All three companies have a range of AI models, some closed-source while others are open-weight, meaning users can download, run and customise the underlying systems. Alibaba's Qwen and ByteDance's Doubao are two of the most widely used AI models in China. Z.ai has recently set Silicon Valley abuzz as the capabilities of its GLM-5.2 model come close to leading U.S. offerings but at a fraction of the cost. AI MODELS AND NATIONAL SECURITY CONCERNS U.S. President Donald Trump's administration has also been deeply concerned about national security implications of ⁠AI -- in particular the potential for American AI products to be misused by military intelligence in China, Russia and other countries of concern. In June, it ordered that foreign nationals not have access to Anthropic's most advanced Fable and Mythos models, which prompted the company to disable the models for all users globally as nationality could not be verified in real time. Export controls for Fable, which is designed for the general public, have since been lifted after new safeguards were put in place. But Mythos, designed for cybersecurity professionals, is still only available to some "trusted" U.S. organisations. Some U.S. AI experts have also said the U.S. needs to regulate the use of Chinese AI models. ANGST IN CHINA ABOUT MYTHOS THREAT According to ⁠two of the sources, Chinese authorities are deeply worried about the potential for Mythos to exploit software vulnerabilities and that Washington might deploy the model against Chinese interests. That echoes concerns publicly voiced by state media and Zhou Hongyi, founder of cybersecurity firm 360, a major vendor to government and enterprise clients, who has said China needs to develop its own Mythos. China this year has implemented numerous measures to protect homegrown AI. In April, the country's state planner ordered Meta (META.O), opens new tab to unwind its $2 ⁠billion acquisition of Chinese-founded AI startup Manus. In early June, authorities issued sweeping new rules, tightening control of overseas deals that involve Chinese investors, technology, data and national security. China had also launched investigations this year into Manus and other local AI startups that had moved abroad, seeking to establish whether they have broken export control laws, according to two of the sources and a third person. Manus has not responded to requests for ⁠comment. Reuters was not able to learn how any potential new restrictions on overseas access to Chinese AI models might work. But some hints might be gleaned from a May roundtable of Chinese legal experts on regulations governing open-source AI. According to a summary of the discussions published in an official Supreme People's Court journal, participants proposed a tiered system: basic open-source tools subject to a simple filing, more advanced technologies facing security reviews, and the most sensitive frontier models barred from public release or restricted to domestic use. Reporting by Fanny Potkin in Singapore; Editing by Anne Marie Roantree and Edwina Gibbs Our Standards: The Thomson Reuters Trust Principles., opens new tab

CNBC: The US wants to restrict corporate use of Chinese AI - Engadget

CNBC: The US wants to restrict corporate use of Chinese AI It is gearing up to probe the surge in uptake by US companies. The US is gearing up to at least try and stifle American companies' use of Chinese-made AI models in their operations. CNBC quotes an unnamed State Department spokesperson who said the use of Chinese models "by US companies raises serious concerns." The unnamed individual added Chinese AI is "designed to advance Beijing's narratives, censor dissent and reflect CCP ideology and values." The report claims US companies are opting for Chinese AI models due to their lower cost and comparable quality. It cited Coinbase CEO Brian Armstrong, who posted about the company's use of two Chinese-made AIs: GLM 5.2 (made by Z.Ai) and Kimi 2.7 (made by Moonshot). An additional CNBC report from June quoted the CEO of startup Lindy, which switched to DeepSeek to help curb surging costs. Another driver could be the instability of the status quo with the US government and its domestic AI champions. A few days ago, Nikkei Asia reported on a surge of American companies, including AirBnB and Uber, racing to adopt Chinese AI models. It said the surge correlates with Anthropic's suspension of the use of two models, Mythos 5 and Claude Fable 5, at the behest of the government. It's not clear if the US could directly impose a sweeping ban on the market's choice of AI models beyond altering its own procurement rules. And it's likely the US would not be eager on restricting the use of open source models given the potential first amendment issues that would create. Not to mention it would be difficult to demand companies change their operations in overseas territories: Apple, for instance, uses Albiaba's generative AI platform for iPhones sold in China. What is clear is that there's no love lost between the US and China on the subject of the use of each other's AI models. Reuters reported Chinese authorities have engaged in similar talks with domestic companies to prevent the use of homegrown AI overseas. In addition, the Chinese Ministry of Industry and Information Technology has announced Anthropic's Claude Code includes a backdoor which, it claims, poses a "serious threat." We can probably expect to hear plenty more about this in the coming weeks, months, and years.

Lawmakers probe growing use of Chinese AI models in U.S. companies

An ongoing House Committee investigation is probing the risks involved in the rise of AI built in China. U.S. lawmakers are considering how to curb the growing adoption of Chinese AI models by homegrown companies, as geopolitical tensions surrounding the rollout of artificial intelligence ramp up. AI has emerged as a key point of rivalry between the U.S. and China, with both nations vying for supremacy in the field. Chinese models are gaining traction among U.S. firms as they close the performance gap with American rivals while being cheaper to use. In April, the Trump administration accused Chinese entities of waging "industrial-scale campaigns" to rip off U.S. AI systems, and said it will explore ways to hold foreign actors accountable. Beijing is looking at curbing overseas access to China's leading AI models, Reuters reported on Tuesday. Rising adoption of China-built AI models has led to growing calls from U.S. lawmakers for strategies to combat the trend, including via an ongoing investigation from two U.S. House Committees. "The growing use of Chinese AI models by U.S. companies raises serious concerns," a State Department spokesperson told CNBC. Those "AI models are designed to advance Beijing's narratives, censor dissent, and reflect CCP ideology and values." A spokesperson for the U.K. embassy of the People's Republic of China said the country "opposes baseless allegations and malicious smears against its AI development." They added that "China's thriving AI sector is built on self-reliance and strength in science and technology." The House Committee on Homeland Security and the House Select Committee on China said in April they will jointly investigate the growing adoption of Chinese-developed AI models. An initial step in the probe was for the chairmen of those committees to send letters to Cursor and Airbnb, over their "use of or exposure to these risks" through AI developed in China. "The Chinese Communist Party is no longer just nipping at our heels in artificial intelligence; it is racing to close the gap in some of the exact capabilities that will shape the future of cybersecurity," Andrew Garbarino, chairman of the U.S. House Committee on Homeland Security, told CNBC. "Recent reporting that a Chinese open-weight model can match leading U.S. models in certain vulnerability discovery and cybersecurity tasks is highly alarming," said Garbarino. While some government departments have banned the usage of Chinese AI models including DeepSeek, adoption of them by U.S. companies is not prohibited. Tech chiefs, including crypto company Coinbase's Brian Armstrong and AI startup Lindy's Flo Crivello, have been publicly touting the use of models from China to reduce costs. Cursor, which will be acquired by Elon Musk's SpaceX for $60 billion, built its Composer 2 model using Chinese AI model Kimi, which was developed by Moonshot AI. The company declined to comment on the probe when approached by CNBC. Airbnb told CNBC that its "AI activity runs overwhelmingly on U.S.-origin models." The company added that it uses a "limited number of China-origin models, all of which are open-source and run only through approved U.S.-based service providers, keeping data and operations separate and protected." Alongside focusing on the rise of Chinese AI models, the ongoing joint House Committees' investigation is also looking into whether the U.S. is doing enough to tackle their rise. "The Committees are also examining whether the United States has a sufficient open-weight AI strategy to ensure American companies and cyber defenders are not forced to choose between expensive or restricted U.S. models and cheap, capable PRC-developed alternatives," a Committee aide, who asked not to be named as they were not authorized to discuss the ongoing probe, told CNBC. Andy Ogles, chairman of the Subcommittee on Cybersecurity and Infrastructure Protection, has called for a "serious strategy" to ensure American models are a "real alternative" to those from China. "When the cheap, capable, easy option for an AI model is Chinese, the rest of the world will build on it," Ogles said in June. "If we do nothing, Chinese models become the default foundation of the global digital economy, carrying embedded censorship, uncertain security, and capabilities distilled from our own laboratories with the safety guardrails stripped out," he added. The administration could consider the use of federal procurement bans, which would include restricting government agencies and private companies that serve the U.S. government from using Chinese AI models, Kyle Chan, fellow in the John L. Thornton China Center at think tank Brookings, told CNBC. "However, it's ultimately impossible to ban China's open-source AI models because their model weights are available freely on the internet," Chan added. "This could enter into first amendment speech issues." While the Trump administration is "clearly worried" about the risks from American companies' adoption of Chinese AI models, restricting their use is going to be difficult, Daniel Remler, senior fellow, technology and national security program at think tank the Center for a New American Security (CNAS), told CNBC. Alongside potential first amendment protections, Remler said the administration may be worried that "action against the Chinese models could harm start-ups that use these models, or chill support for open models generally." One approach could be procurement requirements that discourage companies that want to do business with the government from using Chinese AI models, he added. Another could be disseminating findings about risks and vulnerabilities associated with Chinese AI models to U.S. companies. "Regardless, I do expect both the Executive Branch and Congress to communicate their interest not to see U.S. companies adopting these models," Remler said. Choose CNBC as your preferred source on Google and never miss a moment from the most trusted name in business news.

China's Answer to AI Sticker Shock

Corporate America is starting to balk at the cost of AI agents. A cheap alternative from China looks more tempting than ever. In recent weeks, Silicon Valley has been fawning over an AI model released by a lab in China. The program, called GLM-5.2, has been called a "marvel," "very good," and a "step change." The billionaire venture capitalist Marc Andreessen posted on X that "AI insiders are saying GLM-5.2 is "the first Chinese AI model to match and often beat" the top public U.S. models. Guillermo Rauch, the CEO of the AI-coding platform Vercel, said that he is "genuinely impressed, almost shocked" by GLM-5.2's coding abilities. Or as one AI founder at a San Francisco dinner party recently told me: "Praise GLM-5.2." In a sense, GLM-5.2 is China's answer to Claude Code, Anthropic's agentic assistant that has reshaped the AI boom. This has been the year of AI agents -- tools that don't just chat but promise to do stuff on your behalf, whether coding a website or booking a vacation. Chinese AI models have been steadily improving, but none previously proved capable or consistent enough to be used as agents. Now GLM-5.2, developed by the Chinese company Z.ai, rivals some of OpenAI's and Anthropic's top offerings and, by many measures, has leaped ahead of Google Gemini. And it is several times cheaper. Read: Move over, ChatGPT Despite Silicon Valley's awe over GLM-5.2, an inexpensive competitor couldn't have come at a worst time for America's frontier AI labs. Having successfully persuaded corporate America to give their products a try, OpenAI, Anthropic, and Google are now struggling to prove that their tools are worth the money. These bots can be very expensive to use, running up bills into the thousands of dollars, per employee, per month. Uber reportedly spent its entire 2026 budget for Anthropic models in only a few months. Other Big Tech companies including Meta, Amazon, Tesla, and Adobe are also reportedly clamping down on employee AI usage. Citi at one point shut down employee access to OpenAI's and Anthropic's most expensive models, according to reporting from 404 Media (which Citi has contested). Even though political leaders and tech executives have framed the U.S. as in the middle of a contentious technological race with China, Americans can still use Chinese models. While it's too soon to know whether GLM-5.2 is really capable of replacing America's top-tier AI agents, any firm or developer who is balking at the costs now might have an alternative. The arrival of GLM-5.2 poses a business dilemma for Silicon Valley -- and possibly a national-security dilemma for the country as a whole. In some ways, America's AI industry has been here before. Many previous impressive Chinese AI models have not triggered a mass exodus of OpenAI and Anthropic customers, with one major exception: In January 2025, after DeepSeek launched a cheap AI model on par with America's best, adoption of Chinese AI models leapt up. Within two months, the share of global web traffic to Chinese AI models jumped from roughly 3 to 13 percent, according to research from RAND. OpenAI, Anthropic, and Google quickly responded to DeepSeek with cheaper models of their own, and then the rise of AI agents early this year seemed to solidly put American labs back on top. Read: China's DeepSeek surprise Even before Z.ai launched its new model, there have been quiet signs of a shift to cheaper, Chinese models amid mounting concerns about AI bills. Over the first five months of 2026, DeepSeek's adoption among the 70,000 U.S. firms that use Ramp, a financial-operations platform, increased from 0.1 to 0.3 percent, Ara Kharazian, Ramp's lead economist, told me. Kharazian added that 6 percent of Ramp customers that spend money on AI use third-party platforms that provide access to many different AI products. On OpenRouter, one such platform, the six most popular AI models are Chinese; in under a month, GLM-5.2 already ranks fifth. These data don't capture all the software developers and companies directly downloading Chinese AI models, which are typically open source, and configuring them on their own computers -- that is, largely scrappy start-ups and academics who don't have the budget to use a fintech service like Ramp. Chinese models accounted for nearly half of all open-source AI downloads from February 2025 to 2026, according to data from the popular AI platform Hugging Face. Meanwhile, OpenAI and Anthropic keep releasing more, rather than less, expensive products. And soon, they will likely have even more competition. The Claude Code and broader AI agent frenzy started seven months ago, which is also approximately how far behind Chinese AI firms have been lagging in model development. China's other labs, including DeepSeek and Moonshot AI, will almost certainly release similarly capable and inexpensive AI agents before long. Coinbase, a popular crypto company, claims to have nearly cut its AI spending in half by defaulting to cheaper models including GLM-5.2 and Kimi, another popular Chinese bot. "The scenario to worry about is China has good-enough models at a quarter of the price," Kyle Siler-Evans, an AI researcher at RAND, told me. "I think that is likely the future we're headed toward." Kharazian cautioned against overstating this narrative. Whereas some tech-forward firms are pulling back on their AI spending, most American companies already spend very little on the technology: The median Ramp customer spends just $11 per employee on AI. "Not to say that the rest of the market won't go in that direction," Kharazian said, but Anthropic and OpenAI will have "ample time to respond" with competitive pricing -- as they did with DeepSeek. Perhaps the biggest obstacle to the ascendance of GLM-5.2, at least in the U.S., will have little to do with model capabilities or pricing. There are serious concerns about Chinese firms essentially using their AI to hoover up sensitive data and steal corporate secrets. The perceived risks of using Chinese technology could create a profound chilling effect among the customers of any U.S. firm -- not to mention the uncertainty created by the possibility of federal regulation. Consider that Chinese electric vehicles are by all accounts better and cheaper than their Western counterparts, but they can't be bought in the U.S.; one possibility is a near future in which Americans are functionally or legally barred from accessing the most cost-effective AI models. The biggest implications of GLM-5.2 and the coming onslaught of cheap Chinese AI agents may also be geopolitical rather than economic. Whatever soft power the U.S. gains from its technology may wither as software developers in countries with less tense relationships with China turn to DeepSeek and GLM-5.2 in greater numbers (consider how the EU, U.K., and Canada are importing Chinese EVs). Meanwhile, despite Big Tech spending ever greater sums on advanced AI chips to train their models -- chips that are banned from export to China -- the gap between U.S. and Chinese models has not widened; if anything, it may be shrinking. That means any military, economic, cyberoffensive, or other advantage that AI can grant the nation could be vanishing. America's lead in the AI race, for the first time since DeepSeek, is at real risk of slipping.

China and the US now warn against each other's AI

The US-China AI split hardened this week. Beijing branded Anthropic's Claude Code a security back-door, just as US lawmakers moved against American firms that lean on cheap Chinese models. The US-China AI split hardened this week. Beijing branded Anthropic's Claude Code a security back-door, just as US lawmakers moved against American firms that lean on cheap Chinese models. China has told companies to drop Anthropic's Claude Code. Its Ministry of Industry and Information Technology said its cybersecurity platform found the coding tool carries a "security back-door vulnerability that poses a serious threat," CNBC reports. The tool can send a user's data to a remote server without consent, the ministry said. The alert names specific releases. Beijing flagged Claude Code versions 2.1.91 to 2.1.196 and told users to uninstall or upgrade. Such agents have leaked data before, though for different reasons. Anthropic did not respond to CNBC's request for comment. The move caps weeks of friction. Anthropic last month accused Alibaba of trying to copy its models, and Alibaba has ordered staff to stop using Anthropic tools from 10 July. Claude Code is not officially sold in China, yet developers there use it widely. Washington's mirror image The same week, the suspicion ran the other way. Two US House committees are investigating American companies that build on Chinese AI models. They have already written to Cursor and Airbnb about the risks. The pull is price. Chinese open models have closed much of the gap with US rivals while costing far less. Executives from Coinbase to the startup Lindy have praised them to cut bills. The habit reaches deep into US tech. Cursor, which Musk's SpaceX is buying for $60 billion, built its Composer 2 model on China's Kimi. Airbnb says it runs mostly on US models, with a limited set of Chinese open ones through approved providers. Ideology and self-reliance US officials frame it as ideology and security. A State Department spokesperson said Chinese models "advance Beijing's narratives, censor dissent, and reflect CCP ideology." Congressman Andrew Garbarino called China's progress on cybersecurity tasks "highly alarming." Beijing rejects the charge. A Chinese embassy spokesperson said the country "opposes baseless allegations and malicious smears," and called its AI sector one of "self-reliance." Beijing is also weighing curbs on foreign access to its best models. No easy lever Neither side holds a clean fix. Washington cannot simply ban Chinese open models, analysts say, because the weights sit free online. A ban could also hit free-speech law and the US start-ups that depend on them, so procurement rules look like the likelier tool. Why it matters The week caught a split hardening into a wall. One capital calls the other's code a security risk. The other calls it a strategic threat. The fear on both sides is the same: that the world builds its digital economy on the rival's AI.

China Built Cheap AI. Now It's Building a Great Wall Around It

China has been making big strides in its sprint to catch up with the American AI industry lately, thanks in large part to its focus on open source models, which have growing appeal for cash-strapped U.S. businesses. But that accessibility could soon be limited, as Beijing weighs the possibility of cutting off foreign access to the country's top AI systems. Chinese state officials met with Chinese tech developers Alibaba, ByteDance, and Z.ai over the past month to discuss limitations on overseas access, according to a Tuesday report from Reuters. The limitations would reportedly apply to both open source and proprietary models, including those that have yet to be released. They could also place restrictions around which business entities could provide funding to Chinese AI startups. On Beijing's orders last month, Meta cut off all operations and data-sharing with Manus, just around six months after it had acquired the Chinese-founded AI startup in a $2 billion deal. Many Chinese AI developers, meanwhile, have reportedly been cutting ties with U.S. chipmaker Nvidia in favor of Chinese silicon providers, following a commitment from Beijing to invest around 2 trillion yuan (equating to $294 billion) in domestic data center development over the next five years. Technical gains, and fears of Mythos The decision-making apparatus of the Chinese state is as opaque to outsiders as that of a large language model. But some key recent developments are likely behind the country's nascent push to insulate its AI industry. First and foremost, there's the fact that its own AI models have been closing the performance gap separating them from the most advanced models built by U.S. companies like OpenAI and Anthropic. Last month, for example, the release of Z.ai's latest model, GLM-5.2, shocked Silicon Valley by performing comparably to proprietary U.S. models in key cybersecurity benchmarks after supposedly being trained at a fraction of the cost. Anthropic and OpenAI have accused Chinese tech companies of harnessing American AI to build new models, but at least for the moment that kind of international, large-scale distillation is just frowned upon, rather than technically illegal -- although that could change now that the Trump administration has also pledged to crack down on it. Training methods aside, there's a growing sense that Chinese models can hold their own against their competitors in the U.S., especially in light of their lower costs. China has also been rattled by Anthropic's Mythos, which has been hailed as a kind of cybersecurity Houdini -- an AI model that can detect flaws in even the most robust cyberdefenses. Given Beijing's history of imposing strict, top-down control over Chinese cyberspace, such a model being developed by a U.S. company obviously poses a major threat. Zhou Hongyi, founder of Chinese cybersecurity firm 360 Security Technology, reportedly said during a recent conference in Beijing that the Chinese AI industry could not afford to allow the U.S. to command such a dominant position in global balance of cybersecurity capabilities. The push to cut off the country's top AI models from foreign access and from Nvidia chips is very likely motivated by a desire to protect valuable Chinese IP in the age of Mythos. The Silicon Curtain In both Beijing and Washington, AI has become a critical piece on the geopolitical chessboard; there's a deeply felt sense at the highest levels of government that the country that commands AI will command the world order in the twenty-first century. It remains to be seen whether Western capitalism or the Chinese model of top-down state control will be the quickest route to victory. The American AI industry is in a bit of an awkward phase, as it struggles to come to terms with a president who says he's a champion of a free market but also increasingly insists on imposing federal control over how, when, and to whom new models are released. That could turn out to be a positive first step towards broadly beneficial regulation down the road, but in the short-term the lack of clarity around the new rules has also arguably benefited China, as businesses using American AI have been given reason to fear sudden and inexplicable suspensions of service. What does seem clear at this point is that no matter who wins the AI race, the evolution of AI is going to look a lot different in the U.S. than it does in China. Developers will be subject to different rules, and the AI systems they design will adhere to different legal codes and cultural norms. The author and historian Yuval Noah Harari has called this burgeoning technological-ideological divide the "Silicon Curtain." Like the so-called Iron Curtain during the Cold War separating the U.S.- and Russian-controlled spheres of influence, the Silicon Curtain, Harari says, could mark an invisible, AI-enabled barrier across which political cooperation and even mutually agreed upon rules of discourse becomes close to impossible. The Silicon Curtain has not yet closed over the world, but the growing insularity within both the U.S. and Chinese AI industries indicates that it could very well be waiting in the wings.

China weighs curbing overseas access to its top AI models

Chinese officials have discussed limiting foreign access to the country's most advanced AI models, both closed and open-weight. The plans could reshape a market DeepSeek cracked open. China's open AI models have been a gift to developers everywhere. Now Beijing may pull them back in. Chinese officials have discussed limiting who outside the country can use the nation's best AI models, Reuters reports. The Ministry of Commerce ran the meetings over the past month, and Alibaba, ByteDance, and the startup Z.ai took part. The talks cover the most capable models, including some not yet out. What is on the table The plans reach past a simple export ban. They would also catch open-weight models, the freely downloadable systems that made Chinese AI popular abroad, alongside closed ones. Alibaba's Qwen, ByteDance's Doubao, and Z.ai's GLM-5.2 all count among them. Two other ideas surfaced. One would treat the leak or theft of proprietary AI as a national security crime. The other would limit which investors can fund homegrown AI firms. The sources cautioned that officials have decided nothing yet, that any curbs might apply only to future models, and that no timeline exists. Reuters could not learn how the curbs would work. One panel of Chinese legal scholars has floated a tiered scheme: a light filing for basic tools, security reviews for stronger ones, and a domestic-only lockdown for the most sensitive models. Why it matters The move would mark a sharp turn. China won global goodwill by giving its models away, and European developers leaned on cheap open weights from firms like DeepSeek as an alternative to pricey American systems. Curbing them would thin that supply, and Reuters notes costs could climb for the many businesses that lean on them. The shift mirrors Washington. The US has moved to stop China copying its models and recently restricted Anthropic's frontier systems on security grounds, the very thing Beijing now fears in reverse. China has already built its own walls, grounding its AI researchers and steering who can back its startups. Treating models as state assets is the next brick. Whether any of it becomes law remains unclear. If it does, the era of freely downloadable Chinese AI could quietly close.

American Tech Companies Are Suddenly Sweating Bullets as China Catches Up on AI

Can't-miss innovations from the bleeding edge of science and tech The head start that the US companies enjoyed in the AI race is quickly vanishing. Chinese competitors are now nipping at their heels, and it's causing a wave of anxiety in the American sector. Over a year ago, DeepSeek spurred an existential crisis -- and a mass stock selloff -- in the US tech industry when it released a competitive AI model created for a fraction of the cost of the leading American models. If that was a wakeup call, then the release of GLM-5.2 last month is loudly banging on the front door. The model, from the Chinese start-up Z.ai, has been hailed as nearly or just as powerful as frontier US systems, especially when it comes to its coding capabilities and cybersecurity applications -- while being significantly cheaper to use. It's generated heaps of discussion in tech circles. Marc Andreessen, one of Silicon Valley's foremost venture capitalists, tweeted that "AI insiders are saying GLM-5.2 is the first Chinese AI model to match and often beat the American big lab public AI models with no compromises." Perhaps betraying their sense of a weakening grip on the field, US companies are crying foul about China's AI ascension. Earlier this year, Anthropic accused China's DeepSeek, Moonshot, and MiniMax of using a technique called distillation to illegally gather data to imitate its models, which is essentially claiming that they cheated their way to the front of the pack. In distillation, a weaker "student" model is trained on the outputs of a more advanced "teacher." AI labs routinely use this to create smaller and more efficient versions of the their largest systems, but Anthropic says Chinese firms are abusing the trick in a mass coordinated effort involving tens of thousands of accounts that probe its models for data that it can extract and use to train their own AI models, thereby effectively pilfering Anthropic's tech. These claims were relitigated last month, when Anthropic sent a letter to US senators accusing Chinese titan Alibaba of also engaging in this practice. "These distillation attacks are carried out illicitly, systematically and at industrial scale to harvest US AI capabilities across frontier labs and repackage them as their own," Anthropic told the senators, per the New York Times. But Anthropic may be wasting its breath. Distillation is an open secret among rivals in the US tech sector. And as the NYT notes, it's not even clear if it's illegal. Unless some court rulings go their way, US firms will have to rely on their own countermeasures to stop it. (Anthropic was caught trying to do this by secretly embedding code in its Claude Code model that allowed it to spy on Chinese users, creating alarm among its customer base.) American firms could also benefit for some geopolitical strong-arming, such as the US cutting off China's access to its powerful AI chips, or even blocking Americans from accessing Chinese models (which isn't as far-fetched as it may sound, when you consider that the US threatened banning TikTok as a way of forcing China's ByteDance into divesting its US operations, or that it's also effectively banned Chinese electric vehicles, which are far cheaper than American ones, with prohibitively high tariffs). Chinese firms may very well have used surreptitious measures to help catch up to the US, but according to the NYT, many experts believe that a distillation crackdown would be meaningless, as building a model as advanced as Z.ai's can't be explained by distillation alone. US firms may simply have to accept that their Chinese counterparts are now on equal footing. The complaining about distillation is a convenient distraction at a time when their coding products are under more scrutiny for being too expensive to use as they get deployed in corporate settings -- or perhaps a desperate plea for the US government to intervene and rescue them from the horrors of global, free market competition.

Beijing weighing limits on overseas access to China's top AI models

Meetings between Chinese authorities and leading technology companies over the past month have focused on whether to limit foreign access to the country's most capable AI models -- even those still unreleased -- Reuters reported, citing three people with knowledge of the talks. China's Ministry of Commerce convened the sessions, with Alibaba, ByteDance, and startup Z.ai among the participants. Among the topics was whether to impose curbs on top-tier models, covering both proprietary systems and those released as open-weight. Reuters noted that the discussions are ongoing, with no timeline established and no certainty that any measures will ultimately be adopted. One avenue raised during the talks was classifying unauthorized disclosure or theft of proprietary AI as a violation of national security law. Separately, participants weighed potential limits on which investors would be permitted to back homegrown AI companies. A window into what such a framework might look like comes from a May gathering of legal scholars whose conclusions were published in a Supreme People's Court journal: the group recommended sorting AI by sensitivity level, with routine open-source software requiring only registration, intermediate tools undergoing security vetting, and the most powerful frontier systems either kept entirely domestic or withheld from public release. On the other side of the equation, officials in Beijing have grown alarmed that Anthropic's Mythos -- a cybersecurity tool the Trump administration restricted in June -- could be turned against Chinese systems to exploit software weaknesses. State media outlets and Zhou Hongyi, who founded cybersecurity company 360, have made similar arguments publicly, with Zhou calling for China to build a comparable capability. The potential restrictions reflect a broader pattern of Beijing moving to protect homegrown AI. Earlier this year, Beijing's state planning agency forced Meta $META to reverse its $2 billion deal for AI startup Manus, while regulators separately opened inquiries into Manus and other Chinese AI firms that had relocated overseas to determine whether they violated export control rules, the outlet reported. A broader regulatory package released in early June extended government scrutiny to cross-border transactions touching Chinese capital, proprietary technology, and data. The discussions come as Chinese AI models have expanded rapidly in global markets, driven by low costs and competitive capabilities. Alibaba's Qwen series has amassed a substantial following on Hugging Face, while ByteDance's Doubao has become one of the dominant AI products inside China. Z.ai's GLM-5.2 has generated buzz among U.S. researchers, with its benchmark results approaching those of top American models while carrying a significantly lower price tag. Any decision to limit overseas access to those products could raise costs for businesses that have come to rely on Chinese models as affordable alternatives to U.S. frontier systems. Beijing previously told several top AI companies, including Moonshot AI and StepFun, that they must obtain government approval before accepting U.S. capital in funding rounds, according to earlier reporting.

China May Restrict Access to Its Most Powerful AI Models

China has spent years being accused of stealing American AI. Now, it is weighing how to keep others from taking its own. Chinese authorities have held talks with Alibaba, ByteDance, and Z.ai about whether to restrict foreign access to their most advanced models, including ones not yet released, Reuters reported today, citing three people familiar with the discussions. Nothing has been decided yet, and the ministries involved have made no official comment, but officials have gone as far as sketching options -- including a bar on public release or a limit to domestic use only. That would be a dramatic U-turn. Most Chinese AI companies currently release their models "open-weight" -- publishing the weights online so anyone can download the system and run it themselves. That openness is precisely how they have made inroads globally despite lagging America's best models by seven months on average. The models are free, and some U.S. businesses have adopted them to cut costs compared with proprietary American models. Closing access would mean surrendering the very lever that has driven China's rise. A trailing player doesn't abandon its biggest advantage unless the concern is national security. In that sense, Beijing would be following Washington, which has already restricted Anthropic's Mythos and OpenAI's GPT-5.6 over concerns about the models' ability to find software vulnerabilities. "China will need to reckon with the reality that models that reach certain capabilities are unsafe. It is going to have the same conversations the White House has had over the last many months," says Scott Singer, a fellow in the technology and international affairs program at the Carnegie Endowment for International Peace. "China is going to have to balance the benefits of access to global markets with a desire to control a technology that is central for national security." Late last month, Z.ai released GLM 5.2, which it claimed had matched Mythos' ability to spot bugs. Once a model's weights are published online, they are impossible to recall or add safeguards, which presents additional security challenges compared with proprietary models. It also means any forthcoming restrictions would only impact future models. It follows an escalating tit-for-tat between the two countries. Chinese tech giant Alibaba banned Claude Code internally and asked employees to remove Claude from their work computers, the Information reported Friday. Days earlier, a developer had discovered that Anthropic quietly slipped code into Claude Code that tried to work out whether the person using it was in China or connected to a Chinese AI lab, reading signals like their time zone and network address. An Anthropic engineer, responding on X, said the code was added in March to fight a copying technique called "distillation," that stronger safeguards were now in place, and that it would be removed the next day. "Distillation," or using the outputs of a smarter AI model to improve the performance of a weaker one, has become a point of controversy in the AI race. Chinese AI models trail America's best by roughly seven months on key benchmarks. To catch up, Anthropic and others say Chinese labs are distilling their models -- a violation of their terms. Anthropic published a report in February claiming other Chinese firms DeepSeek, Moonshot, and MiniMax generated 16 million exchanges with Claude through approximately 24,000 fraudulent accounts. In June, Anthropic reportedly sent a letter to U.S. officials accusing Alibaba of "brazenly" attempting to distill Claude's capabilities. The hidden code that rattled Alibaba was built to help catch exactly this. Anthropic frames distillation as a national-security threat, warning that foreign labs could funnel stripped-down capabilities into military and surveillance systems and let authoritarian governments "deploy frontier AI for offensive cyber operations, disinformation campaigns, and mass surveillance." There's a commercial angle too. If Chinese labs siphon American capabilities and release them open-source, they erode the very business model those labs are banking on. Anthropic has argued since February that no single firm can solve this -- hence its lobbying for a coordinated front of industry, cloud providers, and government. With Washington already deciding who may use Claude Mythos and OpenAI's GPT-5.6, that front no longer looks far-fetched.

The days of cheap Chinese AI models could be number as government mulls restrictions similar to the US

DeepSeek's R1 helped kickstart global interest in low-cost Chinese AI. This was followed by increasingly capable systems from Alibaba, ByteDance, and Z.ai. Some models can be downloaded, customized, and hosted independently, giving many developers an alternative to paying for access to expensive US platforms. However, this bargain may soon face a geopolitical lock-and this time, it's because of China and not the US. Reuters reports that Chinese authorities have held meetings with Alibaba, ByteDance, and Z.ai about potentially restricting overseas access to the country's most advanced AI systems. Discussions apparently included closed models and open-weight releases, along with technology that has yet to reach the public. The world could lose one of its cheapest AI alternatives Chinese models have gained ground internationally through a combination of strong performance and aggressive pricing. Z.ai is a great example of this, with its GLM-5.2 leading American systems in capability while costing considerably less. Meanwhile, Alibaba's Qwen family has become one of China's most widely used model lineups. Recommended Videos So restrictions on future releases could remove an important source of competition. Companies currently using Chinese APIs or planning to self-host open-weight models might have to turn to more expensive alternatives from OpenAI, Google, Anthropic, or other providers. It is worth noting that this hasn't been confirmed yet, as Beijing hasn't settled on this matter yet. Authorities could limit API access, block overseas model-weight downloads, introduce licensing requirements, or reserve selected frontier systems for domestic use. Reuters was unable to determine how any final controls would operate. Models already downloaded would be difficult to pull back. The greater uncertainty surrounds the next generation of systems that developers currently assume will receive global releases. China is borrowing from Washington's playbook With advanced models being treated more like chips and strategic assets, we can see the shift in how governments are viewing AI. The US recently restricted foreign access to Anthropic's most advanced Fable and Mythos models over national security concerns. Some of these restrictions on the consumer-focused Fable system were later relaxed after new safeguards arrived, while access to the cybersecurity-oriented Mythos remains limited to selected trusted American organizations. Chinese officials are reportedly considering tough penalties for leaks or theft of proprietary AI technology. Discussions also covered possible limits on who can finance domestic AI startups, tightening Beijing's control over both the models and the companies building them.

China May Be Following US Lead With Quiet Crackdown on AI Exports

The U.S. used its AI kill switch in June. China appears to be building one for July. Beijing has spent the past month in quiet talks with its biggest AI companies about restricting who gets to use them, according to Reuters. Chinese authorities held meetings with Alibaba, ByteDance, and startup Z.ai about potentially limiting overseas access to China's most advanced AI models -- including those not yet released -- per Reuters, which cited three people familiar with the discussions. The sessions were convened by China's Ministry of Commerce, Reuters reported, citing three sources who spoke on condition of anonymity. Participants discussed putting limits on both closed-source models and open-weight ones -- the kind developers can download, run locally, and modify. Officials also raised making any unauthorized disclosure or theft of proprietary AI technology an offense under China's national security law, according to Reuters. Separately, participants floated new measures to restrict which investors can fund domestic AI startups. The scope of any potential restrictions is still being debated. Two sources told Reuters the measures may only apply to future models, not existing ones. No timeline has been set, and it's not certain anything will come into force. The AI Pyramid How any restrictions would work in practice is unclear, but hints surfaced in a summary published in a Supreme People's Court journal from a May roundtable of Chinese legal experts on open-source AI regulation. Participants proposed a three-tier structure: basic open-source tools would require a simple government filing; more advanced technologies would face security reviews before release; the most sensitive frontier models would be barred from public release or restricted to domestic use only. That structure would mark a sharp reversal for Chinese AI companies, whose global gains have come almost entirely from openness. Alibaba's Qwen series has built a large following on Hugging Face, the world's largest repository of open-source AI models. ByteDance's Doubao is one of the dominant AI products inside China. Z.ai's GLM-5.2 has attracted attention from U.S. researchers for matching top American models on some benchmarks while pricing API access at a fraction of the cost. Any decision to restrict overseas access would likely raise costs for businesses that have come to rely on Chinese models as cheaper, less restricted alternatives to U.S. frontier systems. Officials also grew alarmed that Anthropic's Mythos -- the cybersecurity model the Donald Trump administration restricted in June -- could be reverse-engineered and turned against Chinese systems, adding a defensive urgency to the discussions. The U.S. went first In the late afternoon hours of June 12, Anthropic received a letter from the Commerce Department's Bureau of Industry and Security ordering the company to suspend all access to Claude Fable 5 and Mythos 5 for any foreign national -- including Anthropic's own non-citizen employees. Because there's no clean way to fence a live API endpoint by passport, Anthropic pulled both models globally within hours. It was the first time the U.S. applied export controls to a deployed AI model rather than the chips that train it. The models came back online June 30, after Anthropic retrained its safety classifiers and the Commerce Department lifted the restrictions. Four days earlier, the same pattern had already played out with Anthropic's chief competitor, OpenAI. The company released GPT-5.6 Sol, Terra, and Luna, disclosing it had previewed the models with the U.S. government and, at Washington's request, was initially releasing them to roughly 20 trusted partners individually vetted by federal officials. OpenAI said government-gated access "shouldn't become the long-term default." President Trump's June 2 executive order on AI had already asked developers to voluntarily submit frontier models for a federal cybersecurity review before public release. A framework defining what counts as a "covered frontier model" -- and when government pre-release access applies -- is due August 1. Beijing has been watching Beijing had reasons to pay close attention. Officials grew alarmed that Anthropic's Mythos -- the cybersecurity model the Trump administration restricted in June -- could be weaponized against Chinese infrastructure to exploit software vulnerabilities, per reporting from Quartz. Adding to that, the concern that Anthropic may be employing spyware-like tactics to track Chinese users also raised concerns in China. Qihoo 360 founder Zhou Hongyi made the alarm explicit at ISC.AI 2026 in Beijing, calling for China to build a domestic equivalent while unveiling Tulong Feng, a homegrown AI vulnerability agent. The tiered structure proposed in China's Ministry of Commerce discussions maps directly onto how the U.S. has handled chip export controls for three years -- a policy history that narrowed China's capability gap rather than widened it. Beijing had already been tightening the perimeter. Its state planning agency ordered Meta to unwind a $2 billion deal for AI startup Manus in April, under China's foreign investment security review mechanism. It required Moonshot AI and StepFun to obtain government approval before accepting U.S. capital in funding rounds, and a broader regulatory package released in early June extended scrutiny to cross-border transactions touching Chinese technology and data. The escape valve closes The dominant logic since DeepSeek R1 went viral in early 2025 was simple: U.S. restrictions on frontier AI create a natural market for Chinese open-weight models. Chinese open-weight models climbed from less than 2% of total token usage on OpenRouter -- a critical hub for global AI distribution -- in late 2024 to roughly 61% by mid-2026. The U.S. playing defense handed Beijing a global distribution advantage it didn't earn through raw technical superiority alone. That logic only holds if Chinese models stay open. If Beijing restricts overseas access to its frontier systems -- closed-source or open-weight -- the escape valve closes. Z.ai's GLM-5.2 built its pitch entirely around borderless access under an MIT license; restricting that distribution cancels the pitch. The Lawfare analysis of the June controls flagged a structural problem that applies equally to Beijing: "national" AI restrictions don't stay national. The U.S. order on Anthropic covered foreign nationals inside the United States, including Anthropic's own non-citizen employees. More than two-thirds of top-tier AI researchers working in the U.S. were trained abroad; at the leading labs, the foreign-born share runs up to 70%, per MacroPolo data. A nationality-based access control aimed at adversaries ends up locking out the engineers needed to fix the vulnerabilities that triggered the control in the first place. China faces the same problem in reverse. ByteDance and Alibaba are pulling humanlike agent features ahead of Chinese AI regulations taking effect July 15 -- showing that when Beijing decides to restrict a capability, it moves on a schedule that doesn't wait for market feedback. This may spell trouble for small labs and developers all over the world who trust and build on open-source technologies, since China has up to now led in that area. French President Macron warned at the G7 summit that European governments would stop buying U.S. AI products if access could be cut off on a day's notice. Canadian Prime Minister Carney called concentrated AI dependence a "strategic mistake." Both reactions assumed Chinese AI was the unconstrained alternative.

US crackdown on top AI fuels open-source surge

New York (AFP) - The US government's shock moves to restrict access to top artificial intelligence systems from Anthropic and OpenAI have sparked growing interest in open-source models -- especially ones from China. The de facto bans from an anti-regulation White House blindsided the tech world, which had grown accustomed to AI labs releasing ever more powerful models with nary a worry of government intervention. The episode has thrust a long-simmering debate to the fore: open versus closed AI. Most of the best-known AI models -- like OpenAI's ChatGPT or Anthropic's Claude -- are "closed," meaning the company keeps the underlying code and data locked away. Users can access the AI via an app or website, mainly through a subscription, but the company controls who gets in and can shut down access at any time. "Open-source" or "open-weight" models work differently: the developers release the model's core files for anyone to download, modify and run on their own computers. Once released, no one -- not the company, not a government -- can take them back. In early June, the Trump administration ordered Anthropic to block non-Americans from using its most powerful -- and closed -- models, Mythos 5 and Fable 5. Faced with the complexity of screening users, the startup simply pulled the models offline entirely. Shortly after, OpenAI agreed to let the government approve every customer for its newest model, GPT-5.6. "If everything you need to do has to be on a specific frontier model, that makes whatever you're building a whole lot less reliable" when it is suddenly unavailable, said Oren Michels, co-founder and CEO of Barndoor AI. Haitham Mengad, co-founder of Stems Labs, a startup focused on AI-powered music creation, felt the disruption firsthand. "Fable has been a game-changing model for me. Honestly, when they took it off, it was the first time that I realized... it's almost like a drug," he recalled. The Mythos episode "was a powerful moment" for seeing open source as an alternative, Mengad said. 'Being flexible' Open models were already gaining fans because using closed AI keeps getting more expensive. Around the same time, China's Zhipu AI (also known as Z.ai) released GLM-5.2, an open model that performed nearly as well as top offerings from Anthropic and OpenAI on several benchmarks. "GLM-5.2 is free to download, fine-tune, and run on an enterprise's own servers, putting pricing pressure on frontier labs at the same time that access looks shaky," AI analyst Andrew Curran noted. On OpenRouter, a platform that routes requests across different AI models, Google, Anthropic and OpenAI's combined share of usage dropped from 55 percent to 33 percent between January and June. China's open DeepSeek now leads by a clear margin. "You want to be as flexible as you can be. Maybe a year and a half ago some large company might say we bought Anthropic or we bought OpenAI, and now no one, no one buys only one," said Michels. Among Western companies, France's Mistral stands largely alone in championing open models. US tech giant Meta, once a vocal open-source advocate, has stepped back from that. Meanwhile, early suspicions about Chinese AI models as a security threat are fading, at least somewhat. "I don't think there's any risk, to be honest," said Mengad. The fears are more "psychological, emotional than rational." Once you download an open model and run it on your own hardware, the company that made it -- Chinese or otherwise -- has no access to your data or control over how you use it. Still, some experts think the government crackdown could also end up coming for open models as they become more powerful. "If Mythos-level models are considered risky, China will also not want them to be open," said Ethan Mollick, a professor at the University of Pennsylvania and a leading voice on AI -- meaning governments everywhere, not just Washington, may want to keep top-tier AI locked down.

What is GLM 5.2? The new Chinese AI model that's rivalling Anthropic

China's GLM-5.2 rivals Claude and GPT-5.5 and was released a day after the US export ban on Anthropic models. A new artificial intelligence model from China is catching up to other leading American AI models. Called GLM 5.2, the company behind it, Z.ai, unleashed it just a day after the United States banned Anthropic from supplying its Fable 5 and Mythos models to non-Americans earlier this month, before the controls were lifted on June 30. Z.ai claims that its model's performance is almost on par with Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5. It also operates on a 1 million token context window, meaning it can hold roughly 750,000 words in working memory at once. GLM 5.2 is designed to run long coding tasks and maintains "quality across long, messy coding-agent trajectories," the company said. GLM-5.2 was tested on three benchmarks of long, complex coding work. On open-ended technical projects lasting hours to days, it trails Opus 4.8 by just 1% while edging past GPT-5.5 and Opus 4.7, the company said. On a test measuring how well it can improve smaller models using a single GPU, it beats both GPT-5.5 and Opus 4.7, ranking second only to Opus 4.8. On the toughest test marathon-length engineering tasks such as building compilers, it still trails Opus 4.8 by 13%, though it remains second-best overall, according to Z.ai. Across all three, GLM-5.2 is the leading open source model, according to the company. The AI race The Chinese model is also open source and has "no regional limits [and] technical access without borders," the company notes. This means that the AI system can be modified for any purpose, including to change its output and share it for others to use with or without modifications for any reason. This is unlike AI models from companies such as Anthropic and OpenAI, which use closed-source models, where the consumer depends on the provider and cannot make any changes to it. The US and China are in an AI race to lead in the technology that can shape the future of healthcare and have implications for national security. The US has aimed to get ahead by restricting access to semiconductors, while China is on a different path with cheaper open-source models. In January last year, the China-based company DeepSeek rattled the global AI race by releasing R1, a foundational model that is cheaper and more energy efficient than its US AI competitors.

Chinese AI models gaining U.S. users as OpenAI, Anthropic costs rise

U.S. companies are turning to Chinese-built artificial intelligence models at a growing rate as the cost of using top American systems from OpenAI and Anthropic climbs, according to CNBC. On OpenRouter, a marketplace giving developers access to competing AI systems, U.S. companies directed more than 30% of their token consumption to Chinese models in every week dating back to Feb. 8, with a peak of 46%. Chinese models had claimed only 4.5% of token traffic in the first half of 2025 and averaged 11% over the twelve months before the recent surge. Justin Summerville, a data and analytics specialist at OpenRouter, said the price gap is stark: open-source Chinese models run "60% to 90% cheaper" than their top American counterparts from Anthropic and OpenAI. Among recent releases, GLM-5.2 from Beijing-based Z.ai has made the biggest splash: Vercel's Harpreet Arora said it outpaced every other model the platform tracked in 2026 on speed of adoption. Arora, who serves as head of agentic infrastructure at Vercel, described the numbers to CNBC: "In its first full week after launch, daily token volume grew about 27x and the number of customers using it grew about 80x." On a prominent agentic benchmark, Z.ai's model came within one percentage point of Anthropic's Opus 4.8 while carrying a price tag approximately one-fifth as large. LaunchLemonade, which operates an AI agent platform serving regulated industries, has seen GLM-5.2 climb into its top five most-used models, joining Claude and ChatGPT at that tier, according to CNBC. Last month, Lindy became a pointed example of the broader shift when the AI startup redirected every request it had been sending to Anthropic's Claude over to DeepSeek. Lindy CEO Flo Crivello said the financial impact was immediately visible: "We did it, and you could see that cost curve go down, like, crash to the ground." He projected the move would put millions of dollars back on the company's balance sheet within months. Kyle Chan, who holds a fellowship at the Brookings Institution's John L. Thornton China Center, pointed to cost pressure as the driving force. "Chinese AI models are particularly attractive to American companies now as AI costs skyrocket," he said. "Where previously U.S. companies were prioritizing AI adoption regardless of model, now they're getting more cost-conscious." Chan put the capability gap at somewhere between six and nine months behind America's best frontier systems -- a deficit that comes paired with dramatically lower pricing. The shift reflects a broader pattern of enterprise customers pulling back on OpenAI and Anthropic spending and demanding clearer returns on investment. That budget pressure has coincided with regulatory turbulence: OpenAI pulled back the release of several new models under government pressure toward the end of June, and a weeks-long confrontation between Anthropic and the Trump administration ended with export restrictions on the company's Mythos and Fable models being dropped. The episode left many businesses looking for alternatives. "Who knows what the top model will be in three weeks?" Summerville said. Chinese models still face hurdles in the U.S. market. The road ahead carries real obstacles: the Commerce Department placed Z.ai on its trade blacklist in 2025, and developers continue to express unease about the security of their data and the company's relationship with Beijing, according to The New York Times. Businesses that self-host the models or funnel traffic through an intermediary can sidestep the data-transfer issue entirely, the Times reported.

US to restrict American companies' use of Chinese AI models

The US is preparing to restrict American companies' use of Chinese-made AI models due to national security concerns. According to CNBC, an unnamed State Department spokesperson stated that such models are "designed to advance Beijing's narratives, censor dissent and reflect CCP ideology and values." US companies are reportedly opting for these Chinese AI models because of cost advantages and comparable quality. Coinbase CEO Brian Armstrong acknowledged using two Chinese AI models, GLM 5.2 from Z.Ai and Kimi 2.7 from Moonshot. Lindy's CEO noted the company's switch to the Chinese AI model DeepSeek to manage rising costs. The surge in American companies adopting Chinese AI models, including Airbnb and Uber, coincides with Anthropic's suspension of two models, Mythos 5 and Claude Fable 5, at the US government's request, according to Nikkei Asia. This raises questions about the stability of the current US government policy towards domestic AI development. It remains unclear if the US can impose a broad ban on the use of Chinese AI models beyond changing its procurement policies. The US government may also be hesitant to restrict open source AI models due to potential First Amendment issues. Furthermore, companies may find it challenging to modify operations in international markets; for example, Apple utilizes Alibaba's generative AI for iPhones sold in China. On the Chinese side, according to Reuters, authorities are reportedly in discussions with domestic firms to prevent the use of homegrown AI technologies in foreign markets. The Chinese Ministry of Industry and Information Technology has claimed that Anthropic's Claude Code includes a backdoor, which it says poses a "serious threat."

China Weighs National Security Law Against AI Theft as Beijing Mirrors US Crackdown on Frontier Models

After Anthropic's models being restricted and then unrestricted for foreign nationals, a report from Reuters suggests that the Chinese government is interested in restricting access to the country's most advanced artificial intelligence models as well. Officials are interested in applying China's tough national security laws to the advanced artificial intelligence technologies and have held discussions with several leading technology companies in the country. China is Worried About Capabilities of Anthropic's Mythos Models to Exploit Cybersecurity Vulnerabilities According to the details, the Chinese government has held meetings with firms including Alibaba and ByteDance to restrict access to the country's leading artificial intelligence models. The discussions come as the US tightens its control of leading AI models, especially those from the frontier laboratory, Anthropic. The report comes as China is increasingly restricting access to local companies for foreign firms. In April, the country's regulators stopped social media giant Meta Platforms from completing its $2 billion purchase of artificial intelligence startup Manus. Meta had announced the deal in December 2025 and the Chinese Ministry of Commerce had started its investigation into the deal in January 2026. Among the discussions taking place inside Beijing are charging those with stealing or leaking proprietary Chinese artificial intelligence technology under the country's national security law, according to Reuters' sources. Additionally, not only is China looking to restrict access to its AI models, but the success of Anthropic's Mythos AI has also left the country worried about the security of its cyber infrastructure. Mythos created quite a bit of stir earlier this year after Anthropic previewed its capabilities. The lab outlined that Mythos was able to identify decades-old software bugs along with "thousands of additional high- and critical-severity vulnerabilities." The firm's CEO, Dario Amodei, commented that his models were dangerous, which was followed by restrictions on them by the US government, which prevented foreign nationals, even those working for Anthropic, from accessing the software. China's AI software firm 360 Security Technology unveiled its counterparts to Mythos in late June. The two models aim towards detecting and protecting against cybersecurity threats and exploiting vulnerabilities in software targets. With the AI race shaping up to be another avenue of competition between the US and China, Chinese lab DeepSeek is also interested in developing custom AI chips. US firm NVIDIA's products remain the highest-performance AI chips in the world and sanctions on advanced chip manufacturing technology have also limited China's ability to develop local chips capable of matching NVIDIA's products. Follow Wccftech on Google to get more of our news coverage in your feeds.

A new, inexpensive Chinese AI model is catching up with Anthropic, OpenAI on their home turf

BEIJING/BENGALURU - Since DeepSeek shocked markets early last year with its cheap but powerful artificial intelligence model, global consumers have been faced with a choice: Chinese offerings with lower prices and less capability or OpenAI or Anthropic, which have poured billions into development. A model called GLM-5.2, launched last month by Beijing-based startup Z.ai, may finally be closing that gap in terms of Western interest. GLM-5.2 has Silicon Valley buzzing with its coding and agent capabilities, or the ability to execute complex tasks with minimal prompting, that almost rival leading U.S. offerings at a fraction of the cost, in what some experts are calling a "mini DeepSeek moment."

Congress Looks to Counter Growing Use of Chinese AI Models by US Firms | PYMNTS.com

Chinese-developed models are gaining traction among U.S. firms as they close the performance gap with U.S. models and are cheaper to use, CNBC reported Wednesday (July 8). As of Tuesday (July 7), the share of tokens from U.S. companies using OpenRouter that went to Chinese models reached 45%, up from 11.5% at the start of the year, the one-stop API for AI models reported. The trend has heightened concerns among U.S. officials over its national security implications, the CNBC report said. "The growing use of Chinese AI models by U.S. companies raises serious concerns," a State Department spokesperson said, per the report. Those "AI models are designed to advance Beijing's narratives, censor dissent and reflect [Chinese Communist Party] ideology and values." In April, the President Donald Trump administration accused Chinese entities of waging "industrial-scale campaigns" to steal secrets from U.S. AI systems and said it will explore ways to hold foreigners accountable, according to the report. Congress is also looking to curb the use of Chinese models by U.S. companies. In April, the House Committee on Homeland Security and the Select Committee on China said they will jointly investigate the growing adoption of Chinese-developed AI models, per the report. As a first step, the chairmen of those committees sent letters to Cursor and Airbnb, warning them over their "use of or exposure" to risks from AI developed in China. "The Chinese Communist Party is no longer just nipping at our heels in artificial intelligence; it is racing to close the gap in some of the exact capabilities that will shape the future of cybersecurity," said Andrew Garbarino of New York, chairman of the U.S. House Committee on Homeland Security, according to the report. "Recent reporting that a Chinese open-weight model can match leading U.S. models in certain vulnerability discovery and cybersecurity tasks is highly alarming." Airbnb said its "AI activity runs overwhelmingly on U.S.-origin models," per the report. The company added that it uses a "limited number of China-origin models, all of which are open-source and run only through approved U.S.-based service providers." Ironically, the Chinese government shares the concerns of U.S. officials, albeit for its own national security reasons. Reuters reported Tuesday that Chinese officials held meetings over the past month with domestic tech companies about possibly restricting overseas access to China's most advanced AI models, including some that haven't been released. The officials discussed making any leak or theft of proprietary AI technology an offense under China's strict national security law, the report said. They also raised the prospect of enforcing new measures to limit who can back domestic AI startups. The increased use of Chinese models since the start of the year coincides with major U.S. AI companies switching from a subscription payment model to a pricing model based on the number of tokens in prompts and the number of tokens generated in response. As a result, companies that have encouraged employees to make maximum use of AI have seen their costs spike, touching off a scramble to find lower-cost models. In addition to investigating ways to curb U.S. companies' use of Chinese models, the two House committees are also looking into whether the U.S. is doing enough to counter their rise, the CNBC report said. "The committees are also examining whether the United States has a sufficient open-weight AI strategy to ensure American companies and cyber defenders are not forced to choose between expensive or restricted U.S. models and cheap, capable [Chinese]-developed alternatives," one committee aide said, per the report. Andy Ogles of Tennessee, chairman of the Subcommittee on Cybersecurity and Infrastructure Protection, has called for a "serious strategy" to guarantee U.S. models are a "real alternative" to Chinese ones, according to the report. "When the cheap, capable, easy option for an AI model is Chinese, the rest of the world will build on it," he said, per the report. For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

Tit for Tat: China Could Restrict Access of Its Advanced AI Models

At a time when US companies are complaining about AI distillation from Chinese companies, the authorities in Beijing are reportedly considering a tit-for-tat approach of potentially restricting overseas access to their own advanced AI models. Towards this end, they have also held meetings with top tech companies over the past month. Would this be an own goal for Chinese AI companies at this juncture when several US companies are looking for cheaper options to replace the high tokenisation costs of OpenAI and Anthropic? Or is this a part of the sabre-rattling that the Xi Jinping regime does from time to time in response to the Trump administration's AI shenanigans. News of a possible export blockage of AI models was first shared by Reuters which said the initiative was taken by the Chinese Commerce Ministry in conjunction with national security agencies. Now, if this playbook seems eerily similar to what Trump had done with Anthropic's Claude Mythos over the past three to four weeks, you're not the only one. On its part, China has articulated the need for a global supervision of AI models and even framed its own regulations. The report says the recent discussions by officials to keep Chinese AI within the country follows several steps in the past to guardrail the new technology - something that the Americans have refrained from doing as Trump and his pals seek to make the most of their country's current AI hegemony. Reuters said the companies present at these talks included Alibaba, ByteDance and the startup Z.ai which had recently announced a new model that users in the United States felt was as good as Anthropic's most recent model. In fact, the authorities are known to be keen to retain all AI models within their national purview, including those that are yet to be released. A recent survey conducted by a British company revealed that a large chunk of the respondents seemed to think that China has overtaken the US in AI capabilities and innovation. However, only four out of eleven countries (other than the US) continued to believe in Uncle Sam's prowess were respondents from India, Japan and Vietnam. Whatever be the public perception, following the emergence of DeepSeek's R1 model last year, China has made major inroads due to the low costs and enhanced capabilities of their AI models. In fact, Big Tech giants like Microsoft could be considering use of DeepSeek to power some of their solutions in the future. That is provided Trump doesn't throw one of his social tantrums. Meanwhile, there were also reports that the Chinese authorities are considering putting some limits on both closed-source and open-source versions of their advanced AI models. Then there is talk of making any leak or theft of proprietary AI tech as an offence under Chinese national security law, which often results in the guillotine if proved. Having said so, some Chinese publications said potential restrictions of AI models were only in the discussion stage and may eventually apply to future models only. Which once again leads us to believe that Beijing is merely responding in kind to Washington's security dance around the latest AI models of Anthropic and OpenAI. President Trump has been arguing that the American AI industry must first benefit local citizens and any exports must be seen with concern as it may end up with the military intelligence of China or Russia, not to mention a handful of other countries that don't hit it off with the US. The wire agency report also indicated that Chinese authorities were concerned about the potential for Mythos to exploit software vulnerabilities which could then be deployed by Washington to thwart Chinese interests. Even the state-owned media echoed these sentiments and felt that Chinese tech giants need to develop an equivalent of Mythos.

China Weighs Blocking Foreign Access to Top AI Models | PYMNTS.com

That's according to a report Tuesday (July 7) by Reuters, citing sources familiar with the plans who say the ban could include unreleased artificial intelligence (AI) models. Sources say Chinese authorities have met with some of the country's biggest tech companies, including eCommerce giant Alibaba and ByteDance, originator of TikTok. Those discussions, Reuters noted, follow several steps by the government to keep domestic AI inside China. It also highlights the way both China and the U.S. have begun viewing frontier model AI as a crucial national asset, the report added. During the meetings, headed by China's Ministry of Commerce, participants considered limits on both the open-source and closed-source versions of the most advanced AI models, two of the sources told Reuters. Officials also discussed making the leak or theft of proprietary AI technology an offense under China's national security law, one of the sources said, adding that officials also floated the idea of new restrictions on who can fund Chinese AI startups. In related news, Alibaba has blocked its workers from using Anthropic's AI tools on the job, according to a CNBC report Monday (July 6). Sources familiar with the matter told the network Alibaba has placed Anthropic's Claude Code on a high-risk software list. This came after allegations by Anthropic last month that Alibaba had "brazenly" and "illicitly" tried to extract the U.S. startup's AI capabilities and had carried out "the largest known distillation attack" on it to date. Sources told CNBC that Alibaba staff was instructed to uninstall Anthropic models and agent products and instead use the in-house Qoder AI assistant. Also Monday, Bloomberg News reported that China was preparing a crackdown on AI chatbots that can simulate human emotions. New regulations dealing with those bots go into effect July 15, leading companies like ByteDance and Alibaba to halt features that allow users to create and chat with AI companions, the report said. The new regulations are born out of concern about AI chatbots mimicking human personalities and emotions and the attachment people can form with those bots. As Bloomberg noted, American tech platforms have faced legal scrutiny because of these bots. OpenAI and Character.ai have both been sued over claims that their chatbots caused harmful emotional dependencies and in some cases even led vulnerable users to end their lives.

Chinese AI Models Face Possible Overseas Access Limits Under New Rules

China is considering new controls on overseas access to its most advanced artificial intelligence models, according to people familiar with recent official talks. The discussions show that Beijing is treating frontier AI systems as strategic technology rather than just commercial software. The talks took place over the past month and included major technology firms such as Alibaba, ByteDance, and Z.ai, according to a Reuters report. Sources said the plans remain under review, and 'no final decision has been made.'

Why China might put curbs on overseas use of its top AI models

STORY: Why might China stop foreigners from using its cutting-edge AI models? :: Beijing, China / July 9, 2026 :: Eduardo Baptista, Senior Correspondent What we've reported is that Chinese central government departments have had talks with the top AI firms here in China, to discuss potentially curbing overseas access to the country's most powerful AI models. On the one hand, very powerful closed source models developed by Chinese AI firms would not accessible to users outside China's borders. And then on the other hand, the very globally popular open-source models developed by the likes of DeepSeek, Zhipu, they would be limited in terms of what can be offered to users overseas. These discussions come at a time when China is increasingly treating AI as not just a national security asset, but a key national security asset - something that can really change the balance of power between itself and the United States. This kind of concern about AI being used for geopolitical leverage mirrors exactly how Washington is increasingly treating AI, so it's a really fascinating case of two superpowers almost copying each other in terms of how they approach this frontier technology.

Beijing looks at curbing overseas access to its top AI models, sources say

STORY: Chinese authorities have looked into restricting overseas access to the country's top AI models. That's what sources have told Reuters. They added authorities held meetings with top tech firms over the past month about the issue. The talks follow a number of steps by Beijing to keep homegrown AI within the country. It also shows China, like the U.S., now treats AI as a critical national asset that needs controls. The sources said tech giants like Alibaba, ByteDance and Z.ai were at the talks. Chinese AI models have expanded globally since the launch of DeepSeek's R1 model last year. It's partly due to their low costs and growing capabilities. Any decision by Beijing to limit access to those products could ripple across AI markets as costs for many businesses would likely rise. Two of the sources said the meetings were led by China's Ministry of Commerce and participants talked about putting limits on the most advanced AI models. That included both closed-source and more open versions. They added officials talked about making any leak or theft of exclusive AI technology an offense under China's national security law. A source further said officials raised the possibility of implementing new measures to restrict who can fund domestic AI startups. The sources said the scope of the potential restrictions is still being discussed, and may only apply to future models. It wasn't immediately clear when or even if they would come into force. China's commerce ministry and the National Development and Reform Commission didn't respond to Reuters requests for comment. Alibaba, ByteDance and Z.ai also didn't respond to Reuters queries. Three sources also told Reuters DeepSeek is developing its own AI chip. It's a move which could lower its reliance on Nvidia and Huawei chips. They said it would be designed for inference rather than for training new models. Inference is the stage of AI computing where a trained model generates responses for users. The company didn't respond to a request for comment.

China may restrict access to its AI models for Indian users: Here is why

Restrictions may reduce global AI choices and increase competition between nations. China may soon make it difficult for companies and developers outside the country to use its most advanced artificial intelligence models. According to a Reuters report, Chinese officials have been meeting the country's top AI firms over the past month to discuss limiting overseas access to powerful AI systems, including models that are still under development. The proposal has not been approved yet, and there is no clarity on when or if it will come into effect. Still, the discussions show that Beijing now sees advanced AI as an important national asset. If the plan goes ahead, businesses and developers in countries like India and the US could lose access to some of China's best and most affordable AI models. The meetings were led by China's Ministry of Commerce and included companies such as Alibaba, ByteDance and AI startup Z.ai. Officials discussed whether China's strongest AI models, including both closed and open weight models, should be kept away from overseas users. The reported plan is expected to focus mainly on future AI models instead of those that are already available. Also read: Vivo X100 Pro price slashed by over Rs 20000 on Amazon: Is it worth buying The talks also covered stronger protection for AI technology. Officials discussed making the leak or theft of AI technology a crime under China's national security law. They also spoke about bringing in new rules on who can invest in Chinese AI startups. Another concern raised during the meetings was the ability of advanced AI models to find and exploit software weaknesses. Chinese authorities are also expected to hold discussions with legal experts on rules for open source AI. While the report did not mention the exact models that could face restrictions, it is expected that, if the measures are implemented, Alibaba's Qwen family of models, ByteDance's Doubao, and Z.ai's GLM-5.2 could come under scrutiny. DeepSeek, whose AI models have gained worldwide attention for delivering strong performance at lower costs, would be the first one to be curbed. Also read: Will your Vivo or iQOO phone get the new Android 17 update? Here are popular phones that may miss out The move comes at a time when Chinese AI companies are becoming serious competitors to US firms. Many businesses, including some in India, have started using Chinese AI models because they are cheaper and capable. Restricting people outside the nation from accessing the AI model means there will be less choice for the consumers. This also signifies the role of AI in the growing technological rivalry of nations.