A new essay published in Foreign Affairs by Christina Knight and Scott Singer argues that the United States and China have both the incentive and the capacity to cooperate on AI safety, if their governments can find the political will to do so. The piece, published on April 7, arrives at a moment when the dominant narrative in Washington and Beijing is one of AI competition and technological decoupling, making its argument for cooperation both timely and deliberately contrarian. Drawing on recent developments in AI governance, arms control theory, and the history of US-Soviet nuclear cooperation, Knight and Singer contend that the risks posed by advanced AI systems are sufficiently serious and sufficiently shared that neither country can afford to address them unilaterally.
The Case for US-China AI Safety Cooperation
The core argument of the Foreign Affairs essay is straightforward: the most dangerous potential failure modes of advanced AI systems, including loss of human control, catastrophic misuse by state or non-state actors, and the destabilizing effects of AI-enabled military capabilities, are not problems that any single country can solve on its own. An AI system that causes a major accident or is used to launch a devastating cyberattack does not respect national borders, and the consequences of such events would be felt globally. In this context, the authors argue, the US-China AI competition is not analogous to the Cold War arms race, it is analogous to the nuclear proliferation problem, which ultimately required cooperation between adversaries to manage effectively.
Knight and Singer point to several areas where US-China AI safety cooperation is already occurring at the technical level, including joint participation in international AI safety research forums and informal exchanges between researchers at leading AI labs in both countries. They argue that these technical-level interactions provide a foundation for more formal cooperation, and that the primary obstacle to progress is not technical but political. Specifically, the reluctance of policymakers in both Washington and Beijing to be seen cooperating with an adversary on an issue that is simultaneously a matter of national security and economic competition.
China’s Growing Public Anxiety About AI and Jobs
The Foreign Affairs essay is published against a backdrop of rapidly shifting public sentiment in China about the social consequences of AI. According to data cited by analyst Matt Sheehan, Chinese public concern about AI’s impact on employment has surged dramatically over the past year, jumping from sixth place to second place in surveys of the issues that Chinese citizens are most worried about. This shift reflects the growing visibility of AI-driven automation in sectors that employ large numbers of Chinese workers, including manufacturing, logistics, customer service, and content creation.
The surge in public anxiety about AI and jobs creates a complex political dynamic for the Chinese government. On one hand, Beijing has staked enormous prestige and resources on the success of its AI development strategy, and any suggestion that AI is causing widespread job losses could undermine public support for that strategy. On the other hand, the government’s own “AI Plus” initiative explicitly aims to integrate AI into every sector of the economy, which inevitably implies some degree of workforce displacement. Managing this tension, promoting AI adoption while addressing its distributional consequences, is likely to become an increasingly important challenge for Chinese policymakers over the coming years.
The Path Forward: Cooperation Without Naivety
Knight and Singer are careful to avoid naive optimism about the prospects for US-China AI cooperation. They acknowledge that the two countries are engaged in a genuine and intense competition for AI leadership, and that many of the most important AI applications, including military systems, surveillance technology, and critical infrastructure, involve direct conflicts of interest that make cooperation difficult or impossible. Their argument is not that the competition should be abandoned, but that it should be managed in a way that reduces the risk of catastrophic outcomes for both countries and for the world.
The essay concludes with a call for the establishment of a formal US-China AI safety dialogue, modeled on the strategic stability talks that the two countries have conducted on nuclear weapons. Such a dialogue would not resolve the underlying competition, but it could create channels of communication that reduce the risk of miscalculation and provide a framework for addressing the most dangerous potential failure modes of advanced AI systems. Whether the political will exists in Washington or Beijing to pursue such a dialogue remains deeply uncertain — but the Foreign Affairs essay makes a compelling case that the cost of not trying may ultimately be very high.
The timing of the piece is significant. It arrives in the same week that OpenAI, Anthropic, and Google announced a joint working group to combat Chinese AI model distillation, a development that illustrates both the intensity of the US-China AI competition and the degree to which it is generating new forms of conflict that existing diplomatic frameworks are not equipped to manage. Knight and Singer’s argument is that the two dynamics, competition and cooperation, are not mutually exclusive and that the most dangerous outcome would be a world in which the competition intensifies without any cooperative mechanisms to manage its most dangerous consequences.