Chinese AI models are increasingly unsettling Silicon Valley. For years, American AI companies believed they held a comfortable lead in the global AI race, expecting China to catch up eventually—but not anytime soon. That confidence is now beginning to fade as Chinese firms continue to release a string of powerful AI models, from DeepSeek and Z.ai's GLM 5.2 to Moonshot AI's latest open-weight model, Kimi K3. The rapid pace of these launches has sparked growing concern within the US tech industry.
The debate has divided Silicon Valley.
On one side, companies such as Anthropic are urging Washington to impose tighter restrictions on Chinese AI, arguing that it poses a threat to America's technological leadership and national security. On the other, firms including Nvidia, Google and Microsoft believe the US should focus on building better AI systems rather than blocking Chinese competitors.
Why are Chinese AI models causing concern?
The anxiety is not simply because models like Kimi K3 are cheaper.
Chinese AI systems are increasingly matching—or even outperforming—leading American models such as Anthropic's Claude and OpenAI's ChatGPT on several benchmarks. More importantly, many of these models are released as open-weight systems.
Unlike proprietary AI models that require paid API access, open-weight models allow developers to download, customise and run them on their own infrastructure. This makes them significantly more affordable and attractive for startups and businesses that may not be able to afford token-based commercial AI services.
It is this widespread accessibility of Chinese open-weight models that has become the centre of the debate.
Anthropic alleges China copied American AI
Anthropic argues that Chinese companies have not caught up through independent innovation alone but by allegedly extracting knowledge from American frontier AI models.
The company claims Chinese AI firms used large-scale "distillation" techniques, where outputs from advanced AI systems are used to train competing models.
"We have identified industrial-scale campaigns by three AI laboratories—DeepSeek, Moonshot, and MiniMax—to illicitly extract Claude's capabilities to improve their own models," Anthropic said.
According to the company, these firms generated more than 16 million interactions with Claude through roughly 24,000 fraudulent accounts, violating Anthropic's terms of service and regional access restrictions.
Anthropic pushes for tighter restrictions
Anthropic has called on the US government to classify frontier AI model weights as critical national security assets. It is also seeking stricter regulations governing API access and legal penalties for foreign firms accused of extracting data from American AI systems.
According to a New York Times report, both Anthropic and OpenAI have privately lobbied the Trump administration to impose tighter restrictions on Chinese open-weight AI models, citing concerns over intellectual property theft and national security.
The Trump administration appears to be taking those concerns seriously.
Treasury Secretary Scott Bessent recently said the administration is considering sanctions against Chinese companies found to have stolen American AI intellectual property, adding that "open source is not open season on American IP."
White House science adviser Michael Kratsios also criticised what he described as "large-scale, covert industrial distillation" of US AI technology.
These comments came after Anthropic alleged that Moonshot AI's Kimi K3 had been trained using data distilled from Claude, although those claims have not been independently verified.
Much of Silicon Valley opposes a ban
Many of the biggest names in the technology industry disagree with restricting Chinese AI.
Instead, they argue that the US should respond by accelerating innovation rather than limiting access to competing models.
Nvidia CEO Jensen Huang has argued that "the world needs both frontier closed models and frontier open models."
Microsoft CEO Satya Nadella has similarly described open-weight AI as "essential to a healthy AI ecosystem."
Google CEO Sundar Pichai has said the company has long benefited from open-source software and supports the broader open-weight AI movement.
Former OpenAI CTO and Thinking Machines founder Mira Murati also defended open-weight AI, arguing that knowledge is inherently distributed across scientists, engineers, businesses and researchers.
"For AI to benefit from distributed knowledge, it must itself be distributed," she said.
Elon Musk, Meta, IBM and Palantir have likewise backed open-weight AI, saying it promotes innovation, reduces costs and enables more companies to compete.
Nearly 200 companies, operating under the newly formed Little Tech Association, have also urged the Trump administration not to restrict Chinese open-weight AI models.
They argued that many American startups rely on affordable open AI models that cannot easily be replaced by expensive proprietary alternatives offered by companies such as OpenAI and Anthropic.
Instead of broad restrictions, they recommended targeted safeguards against intellectual property theft while allowing continued access to AI models already available worldwide.
Anthropic calls out rivals
Anthropic's chief scientist Julian Schrittwieser questioned the stance taken by several large technology companies.
He remarked that if firms such as Nvidia and Microsoft now champion openness, perhaps they should also open-source technologies like CUDA and Windows.
He later clarified that he supports open AI models but questioned why companies that have historically relied on proprietary software were suddenly positioning themselves as advocates of openness.
"I actually think open models can be very useful! But it's interesting how some historically extremely anti-open-source companies are suddenly all in favour of openness," he wrote.
OpenAI adopts a mixed position
OpenAI appears to be taking a more nuanced approach.
While the company has joined Nvidia, Microsoft, Google, Meta and dozens of other firms in signing an industry letter supporting open-weight AI, reports suggest it has also privately urged the Trump administration to impose tighter restrictions specifically on Chinese open-weight models.
The White House seeks a middle ground
For now, the Trump administration appears reluctant to impose a blanket ban on Chinese open-weight AI.
Instead, officials are reportedly considering targeted measures against specific companies accused of threatening US national security or misappropriating American intellectual property, while allowing broader AI innovation to continue.
