The debate over the rapid development of artificial intelligence has taken a new turn with Anthropic CEO Dario Amodei calling for a slowdown in AI development so that safety measures can catch up. He has warned that AI swarms could potentially take over the internet. OpenAI CEO Sam Altman and xAI founder Elon Musk also supported such calls for restraint.
China questions the call to slow AI
In China, the call for restraint has prompted questions about who benefits from slowing down AI. Lai Jiaqi, a commentator, argues that a slowdown proposed by leading US AI companies would not affect all participants equally.
Top US AI companies have the most computing power, capital, and fully developed software ecosystems, putting them at the forefront of the industry. For these companies, slowing down could consolidate their advantage, while for those trying to catch up, it could reduce the opportunity to close the gap. As Lai puts it, when the vehicles are in different positions, applying the brakes can produce different outcomes.
The Chinese discussion places this argument within the wider technological competition between China and the US. It sees the US as having advantages in cutting-edge chips, advanced models and private capital, while China is narrowing the gap through open-weight models, lower-cost deployment, industrial applications and large-scale diffusion.
The two sides also face different constraints: China is limited by access to advanced AI chips, while the US faces constraints related to power, grid connections and data-centre construction. Chinese companies have responded by focusing more on efficiency, including model distillation, sparsity, quantisation and inference optimisation.
An article on CRI Online similarly argues that US AI companies are caught between safety concerns and the need to maintain technological leadership. They may support collective safety measures, but slowing down could give competitors, including China, an advantage. The article also points to weak or fragmented US regulation and concerns that coordinated slowdowns could raise antitrust issues.
These differences in the competitive landscape help explain, in part, why Chinese companies such as DeepSeek, Zhipu and MiniMax continue to release new models, raise capital and expand their computing infrastructure, suggesting that continued development is seen as an opportunity to narrow the gap with US leaders.
Also Read: India wants to lead BRICS without picking a side. Chinese think that won’t last
Who gets to shape AI?
The Chinese discussion moves beyond the pace of AI development to who controls the technology and what it is used for. Yin Zhiguang, a professor at Fudan University in China, argues that technology itself is not the problem; what matters is who controls it, what purpose it serves and whether it benefits people.
Yin contrasts what he portrays as a US model driven by capital, monopoly and military applications with a China-centred approach that uses technology to improve living standards, connect communities and support development. He points to 5G, roads and digital platforms as examples of how technology can reduce geographical and economic barriers, particularly for rural communities.
Dong Jielin, a research associate at Tsinghua University, argues that the AI divide between China and the US could resemble the existing internet division. AI could develop into US, Chinese, and hybrid ecosystems, shaped by models, chips, cloud computing, data rules, government procurement, and security standards. This could determine which AI tools people can use, whether data can cross borders, and where companies can operate.
Dong highlights that the US focuses on national security and market access, while China focuses on model access, export, content, and social control. Meanwhile, Europe emphasises risk classification.
A Chinese discussion also looks at how the two societies can integrate AI. It portrays the US as having a more decentralised regulatory environment, while China has stronger central coordination, allowing policy, capital, talent and infrastructure to be mobilised for large-scale deployment.
The question of control also extends to regulation. A commentary notes that China’s AI governance is moving towards formal judicial rules and state-led accountability, with courts defining responsibility for training data, copyright, personal information and AI-generated harms.
In the US, AI governance could move from industry self-regulation towards federal legal obligations, including requirements for advanced AI developers to address safety risks.
An article by the Hong Kong Economic Journal takes the strategic argument further, calling AI a “digital nuclear bomb”. From this perspective, accelerating AI development is seen as necessary to ensure that no country or company gains an overwhelming advantage. The logic is similar to mutual assured destruction amid geopolitical competition between China and the US.
Even competition between companies makes it difficult to slow down, as any company that stops developing AI allows competitors to establish a strong lead.
The Chinese discussion also considers the US as important to the global reach of Chinese AI. China cannot assume that countries at odds with Washington will automatically become technology partners.
What matters, these arguments suggest, is whether Chinese AI can become part of a country’s domestic industrial ecosystem. Canada, for example, shows that tensions with the US do not necessarily translate into greater adoption of Chinese technology.
The US and China have the most complete AI ecosystems, but remain interdependent: US companies need China’s market and manufacturing, while Chinese companies need US chips, capital and international channels.
From this view, US-China coexistence could reduce resistance to Chinese AI, while deeper decoupling could push more countries towards systems that exclude China.
The Chinese view suggests that the debate over putting the brakes on AI is hardly only about safety. In their view, it also seems to be about who sets the pace, who controls the technology, and who gains from the next stage of AI development.
For China, slowing AI development could go either way: it could give China more time to catch up with the US, or limit its opportunity to narrow the gap.
Sana Hashmi, PhD, is a fellow at the Taiwan-Asia Exchange Foundation. She tweets @sanahashmi1. Views are personal.
(Edited by Maryam Hassan)