OpenAI fears open-weight models. Should the US be concerned?

The launch of Kimi K3, the largest open-weight large language model from Chinese lab Moonshot, has ignited a debate that conflates two distinct issues: the economic strategies of American AI giants and the technological trajectory of LLMs.

Dean W. Ball, OpenAI’s head of strategic futures, suggested that the US government should manufacture regulatory fear and uncertainty around new models, arguing that open-weight architectures inevitably deter capital spending by frontier labs.

Backlash was swift, with tech leaders like Yann LeCun and Martin Casado contending that open software accelerates innovation and can coexist with proprietary projects. Ball subsequently retracted his claims that a regulatory crackdown was the White House’s “best strategy” and that open-weight models necessarily hinder technological progress.

However, Axios reports that the Trump administration is considering banning K3 and other advanced Chinese models at the urging of American frontier labs. Conversely, a Politico report indicates that the Department of Commerce does not plan to take such action in the near future.

The advantage for major AI companies is clear: open-weight models, which can run on independent infrastructure or within enterprises, provide cheaper intelligence than Anthropic’s or OpenAI’s proprietary offerings. If users increasingly spend outside closed labs, the return on frontier labs’ massive training investments diminishes.

This perspective extends beyond OpenAI. “Strong, frontier-caliber open source models will squeeze margins and drive down prices for frontier companies,” Braden Hancock, co-founder of Snorkel AI and a research partner at the Laude Institute, told TechCrunch. “It won’t necessarily reduce AI usage; quite the opposite is likely.”

For those without stakes in Anthropic or OpenAI, this isn’t a problem. AI will continue to proliferate. So, what justifies the government blocking Americans from purchasing products in ostensibly free markets?

Concerns regarding Chinese models vary. One argument involves protecting US data from the Chinese government, similar to the ban on modern Chinese EVs due to data privacy concerns. However, experts believe that open-weight models hosted on US servers are unlikely to leak data back to China, though it remains a possibility.

Another concern is that these models may harbor implicit bias toward the PRC, though the implications for tasks like coding remain unclear.

A third common worry is that Chinese models lack the guardrails mandated by the US government (through an opaque process) to prevent the exploitation of closed systems or the creation of weapons. Yet, these same guardrails may make US companies more vulnerable: David Sacks, a venture capitalist and Trump adviser, has highlighted cases where US firms turn to Chinese LLMs to bypass security restrictions imposed by US frontier models.

The most significant motivation for restricting these models, however, is the fear that China will outpace the US if frontier labs slow their development.

Sam Bresnick, a China-focused research fellow at Georgetown’s Center for Security and Emerging Technologies, argues that the growing importance of AI to US military operations justifies continued investment in frontier labs. Nevertheless, he notes that the entire question is fraught with complexity.

“Why should the weight of the U.S. government be aimed at protecting these companies from competitors that are being locked out from the U.S. market based on their origins?” Bresnick asks.

Advocates for open AI argue that frontier companies are creating a false binary between innovation and closed models.

“The bigger impact of having these open source models come from China is less that they’re sneaking in back doors, and more that they are owning the innovation,” Hancock told TechCrunch. “You end up with, effectively, an expanded workforce on your model. PyTorch became the industry standard because it was open source, and so the whole community could contribute to it rather than just one company, and it grew and grew, and all the rest of the deep learning libraries kind of died in comparison.”

Hancock and other advocates fear that Chinese LLMs will become the center of international research. Already, US graduate programs primarily build on open-weight Chinese models, and Hancock states that half of the papers students study originate from Chinese institutions, as American frontier labs increasingly hesitate to share their work widely.

“Restricting open models wouldn’t make AI safer,” said Clem Delangue, CEO of Hugging Face, a platform for open AI collaboration. “It would simply hide the risks, concentrate power in the hands of a few, and make it harder for the next generation of builders, researchers, academia, non-profits, and governments to participate in making AI safer and more beneficial for all.”

Bresnick suggests that the real way to slow China’s progress is to focus on chip export controls. A better strategy to preserve US AI leadership would be to stop selling Nvidia H200 processors to China. “That,” he says, “could potentially keep us out of this thorny debate about banning open source technologies that huge numbers of US companies want to use.”

Part of the problem lies in the uncertainty surrounding AI economics. “The open business model, the proprietary business model — neither one is figured out. AI companies are struggling to figure out how to make money on their tools, especially as training costs need to go up and up,” Bresnick points out.

The same challenges playing out in the US are also evident in China, where AI companies struggle to generate revenue and access compute power. The Chinese government encourages open releases for policy reasons, despite the difficulties in capitalizing on them.

Some US companies, including Thinking Machines Lab and Nvidia, are attempting to build a business around releasing open models. Hancock notes that Nvidia would benefit more “if there are dozens or hundreds of companies building AI than rather than two or three that are well capitalized enough to make their own chips,” which explains its investment in Nemotron, a collection of open models.

“The main point is the U.S. would be very well served to have its own very capable, much less expensive open models,” Bresnick said. “It just clashes with the approach the frontier labs have taken.”

With additional reporting from Rebecca Bellan.