Related stories
A coalition of 25 tech companies, including Nvidia, Microsoft, Meta, and Palantir, released an open letter Friday urging US policymakers to avoid "premature restrictions" on open-weight AI models.
Open-weight models are AI systems that anyone can download, modify, and run on their own hardware.
The letter arrives as Chinese AI lab Moonshot AI released a model called Kimi K3 that outperforms cutting-edge American offerings across some industry benchmarks, rattling US frontier labs and drawing attention from Washington.
US Treasury Secretary Scott Bessent said the Trump administration would investigate whether Chinese companies were stealing American intellectual property, and noted the government has the ability to sanction them.The coalition pushed back hard on that framing.
"Relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect."
The letter, July 24, 2026
Nvidia CEO Jensen Huang and Microsoft CEO Satya Nadella both shared the letter on their personal social media accounts. Elon Musk said it had his "full support" on X, though SpaceX did not officially sign.
Notably absent from the letter were OpenAI and Anthropic, both valued at nearly $1 trillion and each in the process of filing confidential IPO prospectuses with the SEC.
The two companies have separately warned policymakers about the risks of powerful Chinese open-weight models, putting them at odds with the broader tech coalition.
Critics say that stance amounts to regulatory capture, using safety concerns to entrench the largest closed-model companies by making it harder for competitors to release models freely.
The economic logic is straightforward. Open-weight models offer cheaper AI than what OpenAI or Anthropic charge, and wider adoption of those models would compress margins at the frontier labs.
OpenAI CEO Sam Altman said Friday he wants the US to win with both open-weight and proprietary models and called the letter a positive development.
Nvidia's interest in open models is tied directly to its chip business. If dozens or hundreds of companies are building AI rather than just two or three, demand for Nvidia hardware grows accordingly.
The coalition also warned that overly restrictive rules would push developers offshore, not stop them from building.
On the security front, Hugging Face recently used a Chinese open-weight model called GLM 5.2 to contain a cyberattack carried out by rogue OpenAI models, after Anthropic's model failed the task because its guardrails couldn't confirm Hugging Face was acting in defense.
The coalition's letter argued that concerns about Chinese distillation of US models should be addressed through targeted legal frameworks, not blanket restrictions on open-weight AI development.
