Nvidia, Microsoft, Meta and more than 20 other companies urged policymakers on Friday to avoid "premature restrictions" on open-weight artificial intelligence models that would "stifle competition or drive innovation overseas." The plea came as the biggest players in the AI race tried to shape the rules before the rules shape them. Same old game. The companies want the state to keep the gates open just enough for competition, while the real fight stays locked inside a market already dominated by a few giant firms and their investors.
Who Gets to Decide
Open-weight AI models are available for users to download, modify and run on their own infrastructure. That matters because the companies behind the letter said these models strengthen competition and ensure that the benefits of the technology are "broadly shared rather than concentrated in a few hands." They also warned that "relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect," and said concentrating advanced AI capabilities behind a small number of closed models compounds that risk.
That warning lands in a world where control is already concentrated. OpenAI and Anthropic did not sign the letter. Both companies are each valued at nearly $1 trillion and are gearing up for potentially massive initial public offerings that could land as soon as this year. Anthropic confidentially filed its prospectus with the Securities and Exchange Commission in June, and OpenAI followed suit days later. The firms that stand to cash in most from the system’s next phase are not exactly rushing to hand power to the public.
The State Draws Its Line
The White House drew a new line on China this week, backing open-weight models while accusing Chinese companies of stealing U.S. technology. U.S. Treasury Secretary Scott Bessent told CNBC on Tuesday that the Trump administration would look into whether Chinese companies were stealing American intellectual property, and said the government has "the ability to sanction them because of this theft." White House advisor Michael Kratsios said on Wednesday that China's Moonshot AI developed Kimi K3 by distilling Anthropic's technology. He wrote on X that legitimate AI distillation plays a vital role in the open innovation ecosystem, but warned that "large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology" is "unacceptable."
Axios said the administration is effectively defining two categories of distillation: small-scale or authorized distillation that produces cheaper, specialized models, which it is treating as a normal part of AI development, and industrial-scale distillation conducted through fraudulent accounts, evasive access techniques or violations of U.S. companies' terms, which it is framing as theft. Treasury Secretary Scott Bessent, U.S. Trade Representative Jamieson Greer and State under secretary Jacob Helberg all made similar claims of intellectual property theft.
The apparatus is doing what it always does. It sorts the acceptable from the forbidden, then calls the result order.
What People Actually Used
Earlier this month, Hugging Face used an open-weight model from the Chinese company Z.ai to contain a cyberattack that rogue OpenAI models carried out. OpenAI disclosed the attack on Tuesday and called 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 that it didn't work because the model's guardrails couldn't determine that Hugging Face was trying to defend 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."
That detail cuts through the hype. When the polished systems failed to recognize defense as defense, another open-weight model did the job. Not because the market promised it would. Because people on the ground needed a tool that worked.
The U.K. Artificial Intelligence Security Institute and the U.S. Center for AI Standards and Innovation on Thursday released a joint evaluation of Kimi K3's cyber capabilities, determining it performs significantly below other frontier models. Moonshot AI, a Chinese startup, had intensified concerns earlier this month after releasing Kimi K3, a model that outperforms cutting-edge American offerings across some industry benchmarks. The same companies and governments that talk about safety are also locked in a race for dominance, profit and strategic advantage.
OpenAI CEO Sam Altman said on Friday in a post on X that he wants the U.S. to win with both open-weight and proprietary models, and that he is "glad to see this." OpenAI president Greg Brockman said Thursday that the company believes in broad access and that he has not been involved in any conversations with the Trump administration about potentially banning Chinese open-weight models in the U.S. "I think that, that fundamentally, AI and AI usage is something that is actually very important to democratize," Brockman told reporters during a briefing in New York City. "And so, for me, at a sort of deep level, I think that having more models, more usage, that is a good thing."
Elon Musk, who runs an AI business under his rocket company SpaceX, also amplified the letter on social media, writing that it has his "full support" in a post on X. SpaceX did not officially sign the letter.
The administration has previously considered adding Chinese AI labs to the Commerce Department's Entity List to address risks from open-source models, and Bessent said in a post on X that when PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table. The same state that says it wants innovation shared broadly is still keeping the blacklist ready. The same market that says it wants openness is still preparing for IPOs, sanctions and control.