The proposal to slow the development of artificial intelligence models has brought the debate over security and strategic superiority in the technology rivalry between the United States and China back to the forefront. Anthropic CEO Dario Amodei called for progress to be slowed to reduce security risks, while US President Donald Trump argued that this approach could amount to a retreat in the competition with China.
In an article published this month, Amodei said that if AI systems reached a level at which they could contribute to developing their own successors, researchers’ ability to understand and control those systems could fall behind. Amodei therefore called for reducing the pace at which models’ capabilities are increased and proposed using the time gained carefully.
Amodei calls for joint oversight
At the center of the proposal is the appointment of independent evaluators at leading AI companies. Amodei also called for companies and governments in democratic countries to coordinate and, in the future, negotiate international agreements regulating the pace at which leading AI systems are developed.
OpenAI CEO Sam Altman, Google DeepMind co-founder Demis Hassabis and Elon Musk of SpaceXAI publicly supported some parts of the proposal, despite being executives of companies competing with one another.
Trump argues that a slowdown could benefit China
Donald Trump opposed Washington deliberately slowing its AI development efforts. Saying that the United States was ahead of China in this field, Trump stated that the side that won the competition would gain a decisive advantage.
Trump also questioned why industry executives would seek regulations that would restrict their own companies. In a post on his Truth Social account, he argued that rules enforced too strictly could drive companies in the sector into bankruptcy. Trump described opposition to AI and data centers as a “sick conspiracy” and argued that China would be the only side pleased by it.
Amodei’s approach also acknowledges that slowing US development could give Beijing a decisive technological advantage if similar restrictions were not imposed on China. Amodei said that regulating the pace worldwide would require cooperation with China, but that any agreement should rely on strong verification mechanisms to prevent one party from secretly violating the rules to gain a military advantage.
Beijing takes a cautious approach to the security debate
Chinese Foreign Ministry spokesperson Guo Jiakun criticized statements by US technology executives. Guo particularly criticized calls to maintain restrictions on exports of advanced chips and semiconductor manufacturing equipment to China, saying that the politics of fear, conflict and destructive competition would obstruct global efforts to govern AI.
Guo said that AI was an important technology for humanity’s well-being and stated that all parties should work together to develop the technology openly and inclusively for the benefit of humanity.
An opinion article published in China’s state-backed Global Times described Amodei’s proposal as an attempt to preserve the United States’ technological superiority. The article argued that excluding China from global AI governance while restricting it through technological barriers was a hypocritical and short-sighted approach.
Chinese companies are closing the gap
Although the United States is considered to retain an overall lead in frontier AI, Chinese companies are narrowing the gap in some technology fields. Developers such as DeepSeek, Moonshot AI, Z. Ai and Alibaba have produced more competitive models despite US policies restricting China’s access to advanced chips.
Researchers in China believe that US laboratories are still ahead in the most advanced AI capabilities. Zilan Qian, a researcher at the Oxford China Policy Lab, said that progress in the United States was faster and more advanced, meaning that researchers in US laboratories could face more dangerous developments than their counterparts in China.
The Chinese government is also acknowledging the risks of AI more openly. Chinese President Xi Jinping called for measures to prevent a loss of control. Chinese regulators have incorporated these concerns into the country’s AI safety framework, while Beijing views the technology as a strategic area for economic development and competition with the United States.
Comprehensive agreement appears unlikely in the short term
Atlantic Council China expert Kenton Thibaut said that the United States and China shared some concerns about AI safety, but that significant obstacles remained. China suspects that Washington could use security talks to limit its technological rise. US officials, meanwhile, doubt whether an agreement with Beijing could be reliably verified and implemented.
Thibaut said a comprehensive slowdown agreement was unlikely in the short term. By contrast, measures to prevent AI-assisted biological weapons, common testing standards and mechanisms to prevent uncontrolled AI activity from being mistakenly interpreted as a deliberate state attack could be more realistic areas for cooperation, he said.
The debate is expected to gain importance ahead of a meeting planned for September 24 in Washington between Donald Trump and Chinese President Xi Jinping. AI governance is emerging as one of the issues that could arise as the two countries seek to reduce the risks of advanced AI while preserving the technological advantage they view as critical to their economies and national security.
