AI Policy Debate Excludes China Despite Its Dominance in Models

A global conversation missing half the world

When more than 200 leading economists and AI researchers—including 16 Nobel laureates—signed an open letter in July warning that artificial intelligence could drive economic transformation "larger than the Industrial Revolution," the signatory list told its own story. Four in five came from the United States and Canada, with the rest almost entirely from Europe. Not a single name represented China.

The letter, organized by the Stanford Digital Economy Lab, addressed a global economic transformation. Yet its composition suggested half the AI world doesn't exist—a troubling omission given that China is home to roughly half of all frontier model development and arguably the world's fastest AI adoption rate, according to two Hong Kong University of Science and Technology professors writing in the South China Morning Post.

Two AI ecosystems, one missing from the room

The information gap runs deeper than most Western observers realize. While Western discourse lags, Western businesses have already moved ahead. Companies including DoorDash, Siemens, and Airbnb have adopted Chinese models from DeepSeek, Zhipu AI, and Moonshot AI—sometimes alongside, sometimes instead of American counterparts.

For cost-conscious organizations, open-weight models that run cheaply on local hardware often make the most sense. Today, that choice increasingly means a Chinese model. The recent release of Moonshot AI's Kimi K3, which matches leading U.S. models on many benchmarks, accelerated this trend. Investors grasped the implications immediately: U.S. chip stocks fell on the news.

But the disconnect extends beyond the models themselves to what AI does to jobs, companies, and economic growth. Nearly all data on AI's economic impact comes from what behavioral scientists call WEIRD societies: Western, educated, industrialized, rich, and democratic. The other half of the experiment is running in China at greater speed and scale—and producing different results.

Diverging approaches to AI's workforce impact

While Western firms announce AI-driven layoffs, a Chinese court recently awarded 260,000 yuan (US$38,400) to a worker fired because his employer claimed AI could do his job. The ruling established that while companies may adopt the technology, they cannot leave employees to bear the cost alone.

Public attitudes diverge sharply as well. A recent Ipsos survey found over 80 percent of people in China reported excitement about AI, compared to less than 40 percent in the U.S. and Britain, where data centers face growing resistance.

Why it matters

This isn't academic. Global companies and their supply chains operate across both ecosystems. Diverging rules create opportunities for regulatory arbitrage, with firms shifting AI deployment to wherever regulations suit them best. Washington's brief export ban on Anthropic's frontier models demonstrated what hard decoupling could mean. The Stanford letter shows how it's already happening, perhaps unconsciously, among economists—even as nearly 30 countries, mostly from the developing world, gathered in Shanghai the same week to join a new China-led body for AI cooperation.

Closing channels of information

The absence isn't intentional but structural. Fewer than 2,000 Americans now study in China annually, down from about 11,000 before the pandemic. The U.S.-China Education Trust warns that America faces a shortage of China expertise within a decade. Language compounds the distance: much of China's AI research and nearly all policy debate happen in Chinese, which few Western researchers read.

The professors argue that universities must build neutral research spaces bridging both ecosystems, pairing scholars from both sides to study how the technology reshapes markets, companies, and workers. Every serious effort to measure AI's economic impact should include data and scholars from China.

The economists are right that we must act now. But the first act should be widening the room. Otherwise, as the authors note, there's little point in having the discussion at all.

These details were first reported by Allen Huang and Frederik Anseel in the South China Morning Post.