OpenAI's fear of open-weight models fuels a US policy fight

The capabilities of Kimi K3, the open-weight large language model from Chinese lab Moonshot, have reignited a debate over whether Washington should restrict rival AI models. OpenAI’s head of strategic futures, Dean W. Ball, argued that the US government should find a pretext to create regulatory fear, uncertainty, and distrust around open-weight models, reasoning that they must necessarily deter capital spending by frontier labs like OpenAI. Ball later retracted his claim that a regulatory crackdown was the White House’s “best strategy,” after AI figures including Yann LeCun and Martin Casado pushed back, arguing open and proprietary software can coexist.

Axios reports that the Trump administration is considering banning Kimi K3 and other advanced Chinese models at the urging of American frontier labs. Politico, however, reports the Department of Commerce would not take that step anytime soon.

The stakes for OpenAI and Anthropic are direct: open-weight models running on independent infrastructure offer cheaper intelligence, and if usage shifts away from closed labs, it could squeeze the returns on their massive training investments. Braden Hancock, co-founder of Snorkel AI and a research partner at the Laude Institute, told TechCrunch that strong open-source models will squeeze frontier companies’ margins and prices, without necessarily reducing overall AI usage. Hancock also argues the bigger risk isn’t Chinese models secretly embedding backdoors but China owning open-source innovation, the way PyTorch once made the US the center of deep-learning tooling — a dynamic some US firms, including Thinking Machines Lab and Nvidia, are now trying to compete in directly by releasing their own open models, such as Nvidia’s Nemotron line.

Concerns over Chinese models include possible data leakage to China, an implicit pro-PRC bias, and the absence of US-mandated safety guardrails — though experts say open-weight models run on US servers are unlikely to send data back to China. Sam Bresnick, a China-focused research fellow at Georgetown’s Center for Security and Emerging Technology, questioned the underlying logic: “Why should the weight of the U.S. government be aimed at protecting these these companies from competitors that are being locked out from the U.S. market based on their origins?” He argues that restricting Nvidia H200 chip sales to China, rather than banning open models, would better preserve US AI leadership, adding that neither the open nor the proprietary AI business model has been settled yet as training costs keep rising.

Why it matters

If Washington moves to block Chinese open-weight models at frontier labs’ request, it would shield OpenAI and Anthropic’s pricing power — but critics warn it would also cede ground in the open-source ecosystem that has historically produced US-led standards like PyTorch.