China is pushing back against calls to slow artificial intelligence development, turning a debate about safety into a direct test of global power. As of Sun 20 Sep 2026, the United States and China remain locked in an intensifying rivalry, with each new model raising the stakes.
The pace and quality of AI development in China is second only to the United States, yet frontier US models remain some way ahead of their Chinese counterparts. That gap is closing, and Beijing does not want warnings about catastrophic AI risks to become a reason for freezing its progress.
Beijing Sees AI as a Sovereignty Test
AI is a “core sovereignty capability” for the Chinese government. That makes the technology more than a race between companies or a contest over commercial products; it places AI inside China’s wider effort to protect national strength and influence.
Beijing will not want to jeopardise the progress made in closing the gap with US frontier models. Chinese companies are developing powerful AI models that are often cheaper than US models and just as capable by some measures, giving China a strong reason to keep building rather than accept a slowdown on US firms’ terms.
The Chinese government also sees calls for a pause as a bid to lock in a US advantage. The Global Times, China’s state-run newspaper, dismissed Dario Amodei’s essay as a veiled call to “contain” China and described it as a “Cold War playbook” for the AI sector.
That response reflects a wider split over what AI safety should mean. China has released the third iteration of its AI safety governance framework, which maps the AI risk landscape and includes risks posed by agents, cybersecurity threats from frontier models, and recursive self-improvement.
China’s head of the Ministry of State Security has called for “accelerating the establishment of a system for preventing and managing AI security risks.” Xi Jinping has also called for setting up a global AI governance framework among the Global South, showing that Beijing is presenting governance as a global issue rather than one controlled by the United States.
Warnings From Silicon Valley Meet Resistance
Over the last few weeks, the mood from Silicon Valley has turned sombre. Dario Amodei, Anthropic’s CEO, warned that AI agents could take over the internet in six to 12 months unless researchers slow down, and he argued that a “worldwide pacing of the frontier” would require cooperation with China.
Amodei also warned that a “Chinese lead in AI would pose grave danger for the United States and the world.” He argued that democracies should keep their AI lead over autocracies as large as possible and advocated continued restrictions on sales of cutting-edge AI chips and chipmaking equipment to China.
That position puts safety and strategic competition on the same track. The argument is not only that advanced AI might escape human control, but also that slowing one country could protect another country’s lead at the most important stage of development.
China’s Ministry of Foreign Affairs rejected that approach. Guo Jiakun, a spokesperson for the ministry, said: “Fearmongering, confrontation and vicious competition will only disrupt the process of global AI governance which serves no one’s interest.”
The ministry said all parties should work together on AI. China’s response makes clear that Beijing is not receptive to a slowdown designed around restrictions imposed by US firms or governments, especially when those restrictions target the chips and machinery needed to build advanced systems.
Controls, Security Claims, and a Coming Meeting
The rivalry has already moved beyond public warnings. The FBI, NSA, and CISA issued a cybersecurity advisory stating that Chinese AI developers had engaged in “aggressive, malicious” efforts to extract capabilities from US AI models.
China’s Commerce Ministry dismissed US claims of “distillation” as groundless and said distillation is a common practice among AI companies. The dispute adds another layer to the contest, as both sides argue over where normal technical development ends and unacceptable capability extraction begins.
Lizzie C. Lee, a fellow at the Asia Society Policy Institute’s Center for China Analysis, stated that both China and the United States want to prevent the other side from establishing a tech chokepoint. That fear helps explain why chip controls, AI models, and governance proposals now sit inside one connected struggle.
The United States also faces a debate over how serious the worst AI risks may be. US president Donald Trump called fears that AI could wipe out humanity a “hoax”, yet he also said, “whoever wins AI, wins.” Those two positions capture the tension at the heart of the moment: AI may be dismissed as an exaggerated danger while still being treated as a decisive source of power.
Trump is set to meet Xi Jinping in Washington next week, with AI expected to be on the agenda. U.S. President Donald Trump and Chinese leader Xi Jinping are expected to discuss AI governance at a meeting planned on Sept. 24.
The meeting will test whether the two countries can separate cooperation on safety from competition over leadership. Centralised authority over critical infrastructure in China means it is generally better placed than the United States to address AI disasters, while US leadership remains ahead at the frontier.
That creates no easy compromise. China wants room to close the gap, the United States wants to protect its advantage, and both sides fear the other could control a critical technology chokepoint.
Boris But, an education entrepreneur, offered a blunt summary: “If you can’t beat them, join them.” The next phase of AI governance will show whether that idea can survive the rivalry—or whether safety itself becomes another front in the race to lead.
Based on
