China’s response to US-led proposals on artificial intelligence safety is best read as part of a wider struggle over technological advantage, in which Beijing sees calls to slow development as an effort to preserve American leverage. As the two countries plan discussions on frontier AI safety risks in mid-September 2026, Beijing has criticised calls for the US to block AI advancement as fearmongering and has called for substantive technical engagement rather than politicised discussions.
The prospective talks come as Washington increasingly treats China’s frontier AI development as a strategic threat. The US has expanded sanctions, export controls and other restrictive measures, while Beijing has pursued domestic alternatives in semiconductors and AI. The issue could also be raised at a summit between Presidents Donald Trump and Xi Jinping on September 24, although the technical content of the mid-September discussions and any prospect of a binding agreement remain unknown.
Both governments recognise risks from increasingly capable AI systems. Their disagreement centres on whether safety rules can be separated from the competition over the computing power, models and industrial capacity needed to develop such systems. China views AI safety agreements as potential mechanisms for US technological containment, while the US sees Chinese frontier AI development as a national security concern.
Dario Amodei, the chief executive of Anthropic, has outlined a framework intended to pace the development of frontier AI model capabilities and allow more time to manage mounting safety risks. He has also urged the US to strengthen chip export controls on China and crack down on alleged model distillation by Chinese AI laboratories.
The Global Times characterised Amodei’s argument in stark terms. It said the essay’s true agenda was “to attempt to curb China's AI development through technological barriers and regulatory monopolies, uphold Washington's monopolistic hegemony in cutting-edge technology, and exclude China from the global AI governance system”.
US controls on advanced semiconductors, first introduced in October 2022 and tightened further in 2023, are central to Beijing’s concerns. Since October 2022, the U.S. Commerce Department has prohibited sales of advanced AI chips to China, initially focusing on Nvidia’s A100 and H100 processors. The restrictions have made access to advanced computing hardware a central pressure point in the competition over frontier AI.
China has responded with a strategy designed to reduce reliance on foreign technology that could be exposed to sanctions. In April 2025, Xi called for a nationwide mobilisation to achieve “self-reliance and self-strengthening” and build an “independent and controllable” AI ecosystem. China’s 15th Five-Year Plan (2026-2030) places artificial intelligence at the centre of its industrial strategy.
Investment has followed that direction. An $8.2 billion National AI Fund was established in early 2025. China’s National Integrated Circuit Industry Investment Fund deployed $20 billion in its first phase from 2014, $29.5 billion in its second phase from 2019 and $47.5 billion in its third phase from 2024.
China’s approach also includes publishing the weights behind its leading AI models, a distribution strategy intended to encourage adoption and reduce dependence on foreign intellectual property vulnerable to sanctions. Whether domestic chip alternatives can sustain competitive frontier AI development without access to the latest US semiconductors remains unknown.
Beijing has also built its own safety framework. Under guidance from the Cyberspace Administration of China, or CAC, China included an explicit future loss-of-control scenario in an AI safety framework released in September 2024. The CAC issued an expanded version in September 2025, adding the principle of “trusted application, preventing loss of control”.
In May, China’s cyberspace regulator issued joint guidelines covering agent systems. The guidelines require developers to improve their ability to discover, intervene in, block and recover from improper agent behaviour. Beijing’s foremost concern among AI threats is the risk to the ruling Communist Party, which fears the technology could be weaponised against it.
The planned talks therefore bring together two governments that acknowledge frontier AI risks while holding sharply different views of the purpose of international safeguards. Whether either side will accept limits that could affect its strategic position remains unclear.
Comments (0)