Beijing says the proposed slowdown risks becoming a new barrier designed to protect American dominance, while safety advocates warn that AI capability is advancing faster than control.
By Publisher Ray Carmen
China is pressing ahead with AI research as a dispute grows over whether the world's leading developers should deliberately slow frontier systems.
China has rejected calls from influential American technology leaders to slow the development of advanced artificial intelligence, turning a debate about safety into the latest contest over global power, rules and technological leadership.
The clash followed proposals from leading U.S. artificial-intelligence executives for companies and governments to coordinate the pace of increasingly powerful systems. Their concern is that the ability to test, govern and contain frontier AI may not be keeping pace with what the technology can do.
Beijing's answer was blunt. China's Foreign Ministry accused the American voices behind the proposals of promoting threat narratives, confrontation and malicious competition. State-aligned commentary argued that a pause or managed slowdown could freeze the current hierarchy in place: the United States would retain its advantage in advanced chips and leading models, while countries still closing the gap would be asked to wait.
Safety Debate Meets a Superpower Rivalry
The disagreement cannot be separated from the wider technology struggle between Washington and Beijing. Successive U.S. administrations have restricted China's access to advanced semiconductors and chipmaking tools. China, in response, has invested heavily in domestic processors, more efficient model training and open AI systems that can be adapted at lower cost.
For Chinese policymakers, artificial intelligence is not merely another commercial sector. It is becoming essential infrastructure for industry, science, transport, healthcare, defence and government. Accepting externally designed limits could therefore carry economic and strategic consequences far beyond the laboratory.
That does not make the safety concerns imaginary. Advanced models can amplify cybercrime, misinformation, surveillance and dangerous scientific knowledge. Developers also continue to debate whether future systems could become difficult to supervise or control. The central question is not whether safeguards are needed, but who writes them, who verifies compliance and whether the same rules apply equally to every major power.
Why China Suspects an Attempt to Contain It
Chinese critics see a contradiction in American technology companies urging restraint after building a commanding lead. A slowdown negotiated by today's market leaders could protect established firms, restrict new entrants and convert private safety standards into global rules. The timing also matters: Chinese models have narrowed parts of the performance gap while often being released through open or lower-cost channels.
The American argument is different. Safety advocates say a coordinated pace would not be intended to halt innovation, but to give researchers, regulators and independent evaluators enough time to understand powerful new capabilities before they are deployed at scale. Without cooperation, each country may fear that slowing down alone would hand the advantage to its rival.
The Race Will Continue
For now, a binding international agreement to slow frontier AI appears unlikely. China has little incentive to accept a framework it believes could preserve U.S. dominance, while Washington is equally reluctant to surrender its lead. The result is a classic security dilemma: both sides may recognise the danger of uncontrolled competition, yet neither wants to move first.
The more realistic path may be narrower cooperation - shared incident reporting, red lines around military use, common testing methods and direct communication when an AI failure threatens national security. Such measures would not end the race, but they could reduce the risk that rivalry turns technical failure into an international crisis.
China's refusal is therefore more than a rejection of slower development. It is a declaration that the future rules of artificial intelligence will not be written by one country or one group of companies alone. The race is accelerating, and the struggle to govern it has now become part of the race itself.