China rejects the US case for slowing AI development

Why Beijing sees a slowdown as strategic

China trails the United States in the quality and pace of AI development, according to analysts, but the gap is narrowing. That makes any proposal to coordinate development politically difficult.

Lian Jiai Su of technology research group Omdia said Chinese officials are likely to interpret calls for a slowdown as an attempt to lock in an American advantage. Amodei explicitly connected his proposal to the rivalry between democracies and autocracies, arguing that it is important to preserve as large a lead as possible for democracies.

His proposed measures include maintaining restrictions on exports of advanced chips and equipment to China. Some experts already regard those restrictions as a form of unilateral forced deceleration.

China’s Foreign Ministry responded that “intimidation, confrontation and aggressive competition” would only disrupt global AI governance. That statement does not mean Beijing has dismissed the wider risks described by Amodei. It means the country is unwilling to accept safety as a reason to give up strategic ground.

The contrast is sharper because the safety debate in the United States is itself divided. Amodei has warned about catastrophic loss of control over advanced systems, while US President Donald Trump called fears that AI could destroy humanity a hoax this week.

Trump is expected to discuss AI with Chinese President Xi Jinping when they meet in Washington next week. Analysts doubt the meeting will produce a major breakthrough.

China is building safety rules while pushing deployment

China released the third version of its AI safety governance system this week. Gabriel Wagner, a researcher at Beijing-based consulting firm Concordia AI, described it as a map of AI risks with broad recommendations for responding to them.

The updated framework includes:

Wagner believes the positions of Amodei and the Chinese government may have more in common than their public statements suggest.

China’s approach is built around state control. The Communist Party closely monitors threats to its political authority and seeks to keep new technologies within boundaries set by the government. Su said centralized control over critical infrastructure and commercial activity could give China more ability than the United States to respond to an AI-related catastrophe.

But control from the center does not automatically produce stronger company-level oversight. Lea Wang of the Carnegie Endowment for International Peace (CEIP) said Chinese companies voluntarily monitor and evaluate their systems less often than American companies. She also said neither government appears prepared for the kind of loss-of-control incident experts have warned about in recent weeks.

Chinese regulation is iterative, Wang added: authorities adapt quickly and introduce new measures as risks change.

The country’s algorithm registry has operated since 2022. This year’s warnings about AI agents suggest that the focus is shifting from controlling what AI says to controlling what it does. Beijing has also introduced nationwide rules aimed at limiting emotional dependence on AI companions.

The quiet issue is enforcement. A framework that offers general recommendations, combined with fast-changing rules and limited voluntary evaluation by companies, says less about how a dangerous system would actually be stopped than about which risks officials are watching.

The argument from everyday use

China is also promoting AI across the economy, linking adoption to economic growth in its new five-year plan.

Local governments offer companies funding and low rents. Recently, entrepreneur Boris But was invited to present Hampshire Education to a small city three hours from Shanghai that wants to establish itself as a technology hub.

Industrial companies are being encouraged to move production processes onto AI systems. A representative of Black Lake Technologies, a platform for industrial AI agents that says it works with 40,000 factories worldwide, said safety and controllability remain active topics of discussion. In manufacturing, however, those questions tend to be practical: companies mainly want to know whether an AI system will be reliable.

Black Lake Technologies says it works with 40,000 factories, while the digital medical service A-fu recently reached 150 million users.

A-fu uses chatbot versions modeled on well-known doctors and trained on real clinical research. Its scale illustrates why Chinese policymakers and businesses may treat AI adoption as an economic necessity rather than a distant safety debate.

The same logic is visible in But’s business. About two years ago, the education entrepreneur realized that his company needed to adopt AI or risk being pushed out of the market. Affluent parents seeking to send their children abroad were arriving with information and opinions gathered from chatbots rather than competing companies. But responded by developing his own service for admissions and education support.

My read is that this is less a disagreement over whether advanced AI is risky than over who gets to decide what acceptable risk looks like. China is building controls around AI while making deployment part of its growth strategy. Its users already treat many AI services as ordinary tools, and its companies are being pushed toward industrial adoption.

Ni Tao, founder and editor-in-chief of technology site yangtzeer.com, said most people see AI as a useful everyday tool rather than an abstract threat and therefore prefer to coexist with it. Concerns about jobs and disinformation exist, he added, but they coexist with the belief that China must master AI instead of slowing its development.

That combination is the tension at the center of China’s position: the state wants to govern what AI does, but the economy is being organized around doing more of it. The result is not a rejection of AI safety. It is a demand that safety be designed around continued acceleration, not used to halt it.