Jensen Huang made his first-ever post on X to defend open-weight AI models, warning that government restrictions could backfire. The Nvidia CEO shared a letter signed by 25 companies and organizations — including Microsoft, Meta, and Hugging Face — arguing that open models strengthen security and accelerate innovation. The move comes as US officials consider curbs on Chinese AI models, specifically Moonshot AI's Kimi K3, which launched July 16 with 2.8 trillion parameters.
The Open Model Argument
The letter compares open-weight AI models to open-source software, calling them a key to American leadership. It argues that open models allow many teams to check code and fix flaws quickly, while closed models concentrate risk. The signatories also distinguish distillation — training a model using another's output — from theft, calling it normal research. About a year ago, the US government backed open models in its AI Action Plan, labeling them a strategic asset.
OpenAI and Anthropic did not sign the letter. Both companies have warned that strong Chinese open models pose risks. The split highlights a growing divide in Silicon Valley over how to handle the rapid advance of Chinese AI.
China's AI Push and US Response
Moonshot AI's Kimi K3, with 2.8 trillion parameters, is the latest Chinese model to draw scrutiny. White House adviser Michael Kratsios has claimed that Moonshot copied a US model to build it. The letter's signatories reject that framing, arguing that open models enable sovereignty and let countries build their own AI without relying on a few US companies.
Chinese AI models already outpace US rivals in daily use, according to the article. In January 2025, a cheap model from China's DeepSeek caused Nvidia to lose nearly $600 billion in a single day. That event still hangs over the industry as regulators weigh new restrictions.
A Shift in Tone
Huang's first X post is a departure from his usual silence on social media. By wading into the debate, he signals that Nvidia sees open models as critical to its business and to US competitiveness. The letter warns that restricting open models could harm security by concentrating risk in a few closed systems, and could slow innovation by limiting the number of teams that can inspect and improve the code.
The debate is far from settled. US regulators are now considering whether to impose curbs on Chinese AI models like Kimi K3, while the industry remains divided over the best path forward. The next move belongs to Washington.
