As Chinese President Xi Jinping and U.S. President Donald Trump met in Washington, D.C., the United States, on September 23-25, artificial intelligence (AI) has moved visibly toward the center of China-U.S. relations.  

During the visit, President Xi and President Trump had an in-depth exchange of views on AI, pointing out that the two countries should leverage respective strengths rather than guard against each other. The two heads of state agreed to make good use of AI dialogue mechanisms between the two countries and jointly prevent the misuse and abuse of AI. 

As early as in mid-September preparatory talks in New York City, the U.S., between both countries’ economic and trade officials have already included AI-related issues. U.S. officials say the two sides have agreed to establish a formal dialogue, including a possible mechanism for communication around serious AI incidents, with further discussions expected. This is an important signal: AI is no longer only an arena of strategic technological competition between the two countries. It is becoming an area in which some higher-level form of cooperation is increasingly necessary. 

Beyond the race 

The significance of this development should not, however, be misunderstood. China and the U.S. do not face a simple choice between competition and cooperation. AI has become too important to economic development, national security and technological sovereignty for strategic competition to soften. At the same time, AI research and production remain embedded in internationally interconnected scientific, industrial and infrastructural networks, while its consequences are too global for technological separation to offer a sustainable solution. The real challenge is therefore how to govern growing competition in a world of continuing technological interconnection. 

The current AI rivalry is the latest stage of a longer and intensifying technological dispute. For years, information and communication technologies, semiconductors, digital platforms, data infrastructures and advanced manufacturing have become increasingly central to the changing balance of economic power. China’s technological upgrading has reduced forms of dependency that characterized earlier stages of its development, while the U.S. has increasingly treated Chinese advances in strategic technologies as questions of economic and national security. Export controls, industrial policy and efforts to secure critical supply chains have followed. 

Yet describing this simply as a race can obscure as much as it reveals. National technological capabilities matter, but innovation does not take place inside sealed national containers only. Modern AI depends on scientific communities, universities, multinational firms, semiconductor supply chains, cloud infrastructures, natural resources, open-source ecosystems and flows of knowledge that cross borders. Historically, successful technological development has usually combined the construction of domestic capabilities with selective forms of international openness. 

This is particularly important because the language of an “AI race” is not politically neutral. In the United States, competition with China has increasingly become intertwined with arguments for massive investment in data centers, energy and semiconductors, for maintaining regulatory flexibility and for preserving the position of American frontier firms. OpenAI, for example, has explicitly described AI as a race that the United States “can and must win” and has invoked competition with China in arguments about infrastructure and policy. These concerns reflect a real technological rivalry. But race narratives can also become performative; if each side assumes that any restraint will simply allow the other to move ahead, acceleration becomes the apparently rational response for everyone. 

The result can resemble a technological security dilemma. American restrictions and acceleration reinforce Chinese efforts toward self-reliance; Chinese advances then reinforce perceptions in Washington that stronger restrictions and faster development are necessary. Neither side needs to seek uncontrolled escalation for the interaction itself to produce it. This is precisely why institutionalized dialogue matters. Mechanisms for incident communication, exchanges around cyber risks and misuse, and shared understandings of where human control must remain indispensable can reduce the danger that technological competition becomes self-reinforcing and escalates beyond either side’s intentions. 

Different visions 

There is, however, another dimension that deserves additional attention. China and the United States may not simply be competing to build the same technology faster. Their AI ecosystems also reflect different institutional arrangements, development priorities and, to some degree, different social imaginaries of technology. 

A powerful strand of the American frontier-AI debate is organized around the prospect of artificial general intelligence (AGI) or super-intelligence. This framing places questions of autonomy, alignment, loss of human control and even existential risk at the center of discussion. China has its own debates about these dangers and concerns with advanced AI risks. But the dominant emphasis in Chinese policy has often been different. The “AI+” strategy, for instance, presents AI primarily as a general-purpose technology to be integrated into science, industry, consumption, public services, governance and broader economic and social development. 

This contrast should not be exaggerated into a simple opposition between “Western” and “Chinese” approaches or philosophies. Both societies contain competing visions of AI, and the debate is changing rapidly. Still, it raises an important question: What do different societies expect AI to be for? If one ecosystem is strongly oriented toward a technological frontier imagined as AGI, while another places greater emphasis on diffusion across production, infrastructure and public life, they may also prioritize different risks and forms of governance. Cooperation therefore requires more than technical agreements. It requires acknowledging that legitimate differences may exist over the purposes and social organization of AI itself. 

This is also why AI cooperation between China and the U.S. matters far beyond their bilateral relationship. For many countries in the Global South, the central questions include access to infrastructure and knowledge, the capacity to build domestic technological capabilities, the effects of AI on employment and development, and the ability to retain meaningful policy autonomy. 

Greater responsibilities 

The responsibility of the two leading AI powers is therefore larger than avoiding direct confrontation. Their relationship will influence whether the emerging global AI order becomes more fragmented and hierarchical or allows more space for scientific exchange, technological diffusion and different development paths. Cooperation should not mean establishing a closed bilateral condominium over AI governance; nor should it require China and the United States to converge on a single regulatory model or philosophy of technology. It should help create conditions in which competition remains bounded and responsible, shared risks can be addressed, and other societies retain room to develop their own capacities and priorities. 

The most promising signal around the current summit may therefore be modest but important: the recognition that strategic rivalry does not eliminate the need for dialogue. In AI, as in many previous technologies, competition and cooperation will necessarily coexist. The objective should therefore be to prevent the metaphor of a race from crystalizing an undesired future for this technology. If China and the U.S. can develop mechanisms for managing shared risks while maintaining scientific communication, respecting legitimate efforts to develop technological capabilities and leaving space for the autonomy of others, their cooperation could become more than a bilateral stabilizing device. It could help establish a more plural and governable technological order. 

In that sense, the deeper question raised by the summit is whether the world’s leading technological powers can learn to compete and cooperate without allowing technological rivalry to harden into domination or permanent confrontation. 

The author is an associate professor of Sociology at University College Dublin, Ireland, and a part-time associate professor with the Department of Social Sciences at LUT University, Finland 

Copyedited by Elsbeth van Paridon  

Comments to dingying@cicgamericas.com