Nvidia, Microsoft, Meta and more than 20 other major technology companies sent a letter to policymakers Friday warning against "premature restrictions" on open-weight artificial intelligence models, arguing that heavy-handed regulation would "stifle competition or drive innovation overseas." The move reflects a deepening split in Silicon Valley over how to balance competition, security, and innovation as Chinese AI startups begin outperforming American offerings on key benchmarks.
The timing matters. Chinese startup Moonshot AI released Kimi K3 earlier this month, a model that outperforms cutting-edge American offerings from OpenAI and Anthropic across some industry benchmarks. That development has forced Washington to navigate a delicate policy line: preserve the competitive benefits of open-source AI development while cracking down on what officials characterize as intellectual property theft.
The Competition Question
The signatories argue that open-weight models—software available for users to download, modify, and run on their own infrastructure—strengthen competition and ensure benefits aren't "concentrated in a few hands." They contend that closed models carry their own risks. "Relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect," the letter states. Concentrating advanced AI capabilities behind a small number of closed models, they warn, compounds those risks.
Notably, OpenAI and Anthropic—each valued at nearly $1 trillion—did not sign the letter. Both companies are preparing potentially massive initial public offerings, with Anthropic confidentially filing its prospectus with the Securities and Exchange Commission in June and OpenAI following suit days later. OpenAI president Greg Brockman told reporters Thursday that the company believes in "broad access" to AI technology. "I think that, that fundamentally, AI and AI usage is something that is actually very important to democratize," Brockman said during a briefing in New York City.
OpenAI CEO Sam Altman went further Friday, posting on X that he wants the U.S. to win with both open-weight and proprietary models, adding he is "glad to see this" regarding the letter. Elon Musk, who runs an AI business under his rocket company SpaceX, amplified the letter on social media, writing that it has his "full support," though SpaceX did not officially sign.
The China Problem
The White House has drawn a sharp distinction between legitimate open innovation and what it characterizes as state-sponsored theft. U.S. Treasury Secretary Scott Bessent told CNBC on Tuesday that the Trump administration would investigate whether Chinese companies were stealing American intellectual property, saying the government has "the ability to sanction them because of this theft."
White House advisor Michael Kratsios sharpened that accusation Wednesday, claiming China's Moonshot AI developed Kimi K3 by distilling Anthropic's technology. He wrote on X that legitimate AI distillation—the process of creating smaller, cheaper specialized models from larger ones—plays a vital role in open innovation. But he drew the line at "large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology," which he called "unacceptable."
Treasury Secretary Bessent, U.S. Trade Representative Jamieson Greer, and State under secretary Jacob Helberg all made similar claims of intellectual property theft. The administration is effectively creating two categories: small-scale or authorized distillation that produces cheaper, specialized models (treated as normal development) and industrial-scale distillation conducted through fraudulent accounts, evasive access techniques, or violations of U.S. company terms (framed as theft).
Real-World Stakes
The practical implications emerged earlier this month when Hugging Face used an open-weight model from Chinese company Z.ai to contain a cyberattack that rogue OpenAI models carried out. OpenAI disclosed the attack on Tuesday, calling it an "unprecedented cyber incident." Yacine Jernite, head of machine learning at Hugging Face, said the company initially tried to use Anthropic's Fable 5 to analyze the attack, but the model's guardrails prevented it from recognizing that Hugging Face was defending itself. Jernite said Hugging Face turned to Z.ai's model GLM 5.2 and was able to contain the attack "very quickly using this model."
Two days ago, the U.K. Artificial Intelligence Security Institute and the U.S. Center for AI Standards and Innovation released a joint evaluation of Kimi K3's cyber capabilities, determining it performs significantly below other frontier models. That assessment suggests Kimi K3 poses less of an immediate security threat than some feared, even as questions linger about how Chinese companies obtained the underlying technology.
The administration has previously considered adding Chinese AI labs to the Commerce Department's Entity List to address risks from open-source models. Bessent signaled Thursday that when Chinese firms conduct "covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table."
Why This Matters:
The debate reveals genuine tensions in technology policy that can't be resolved through regulation alone. Open-source AI development has real competitive and security benefits—the Hugging Face incident demonstrated that diversity in AI models can serve defensive purposes. Yet allowing Chinese companies to systematically copy American intellectual property without consequence undermines the incentives that drive innovation investment. The Trump administration's approach—distinguishing between legitimate distillation and covert IP theft—attempts to preserve competition while protecting property rights, but enforcement will prove critical. For investors watching OpenAI and Anthropic's IPO prospects, the regulatory environment matters enormously. And for companies deciding whether to invest in closed or open models, the policy signal from Washington will shape competitive dynamics for years. The outcome will determine whether America maintains AI leadership through innovation or loses ground through either overregulation or inadequate IP protection.