Jensen Huang, Nvidia’s CEO, has made his debut post on X, formerly known as Twitter, to champion open-weight AI models and caution the US government against restrictive measures. By open models, Huang refers not to the company OpenAI but to AI models that are freely accessible for download and reuse.
On July 24, Huang shared a letter signed by Nvidia and 24 other organizations including Microsoft, Meta, and Hugging Face. The letter emphasizes open models as essential to maintaining American leadership in AI, likening them to open-source software that underpins much of the internet today.
The timing of this appeal is significant. The US government, in its AI Action Plan released about a year ago, recognized open models as a strategic asset. Currently, however, officials are considering restrictions on Chinese models like Kimi K3, launched recently by Moonshot AI with 2.8 trillion parameters, placing it among the top global AI models.
The Debate Over Openness and Security
The letter outlines safety advantages of open models, highlighting how their transparency enables many teams to audit and fix issues quickly, unlike proprietary models held by few companies, which the letter identifies as potential weak points. It also clarifies that model distillation a process where one AI is trained using another’s output is standard research practice, not piracy.
Meanwhile, tensions simmer as White House advisor Michael Kratsios alleges that Moonshot’s Kimi K3 was developed by copying a US model. Notably absent from the letter’s signatories are OpenAI and Anthropic, who have warned Washington about the risks posed by powerful Chinese open models.
This rift illustrates the ongoing struggle over AI openness. Chinese models already surpass US counterparts in daily usage, adding pressure to the debate. Huang and the co-signers urge policymakers to avoid premature restrictions that might suppress competition or push innovation overseas.
