Silicon Valley divided over open-source AI and China

2 min readUpdated July 29, 2026

Whether Chinese AI models should remain openly accessible or face government restrictions has become one of the most contentious fault lines running through Silicon Valley, pitting major technology companies against one another, according to The New York Times.

In what the Times described as his first-ever post on X $TWTR, Nvidia $NVDA CEO Jensen Huang declared that "the world needs both frontier closed models and frontier open models." Microsoft $MSFT CEO Satya Nadella followed shortly after with his own post calling open-source software "essential to a healthy A.I. ecosystem." An open letter endorsing open models, which both executives signed, also drew support from Meta $META, Palantir $PLTR, and IBM $IBM, according to the Guardian.

Anthropic has taken its concerns about Chinese open-source AI directly to Washington regulators, according to The Times. The company's head of public policy, Sarah Heck, described what she called "illicit, adversarial distillation" as "IP theft and industrial espionage," according to The Times. OpenAI has separately urged the Trump administration to require mandatory security evaluations for some new AI models.

The release of Kimi K3 sharpened the conflict. The open-weight model, built by Chinese startup Moonshot AI, has been described by its developer as capable of matching the top offerings from Anthropic and OpenAI, according to The New York Times. Globally, more developers are pulling from Alibaba's Qwen than any other AI model family, and it has served as the foundation for over 100,000 purpose-built models, according to The Times.

Sanctions against Chinese firms accused of stealing American intellectual property were among the options the administration had weighed, Treasury Secretary Scott Bessent said, according to the Guardian. Rather than a sweeping prohibition, people familiar with the deliberations said U.S. officials were leaning toward handling Chinese open-source models on a case-by-case basis under national security authority, according to The New York Times.

Inside any AI model, "weights" are the underlying numerical parameters shaped during training that govern how the system generates its outputs. Making those weights public allows anyone to download and operate the model locally, customize it, or adapt it for specialized uses, according to The New York Times.

Update, July 29, 2026: Nvidia $NVDA assembled dozens of companies into a new alliance backing open-weight AI tools, according to reports. Prominent developers of closed-weight AI models were notably absent from the alliance. (per NBC News)

Silicon Valley splits over open-source AI as Chinese models gain ground

Related Content

Related Content