Artificial intelligence has moved into the center of the U.S.-China relationship as President Donald Trump and Chinese President Xi Jinping discuss the future of AI governance during Xi's state visit to Washington. The talks come as both countries compete for leadership in advanced AI while researchers, technology executives and government officials increasingly debate how to manage potential risks from increasingly capable systems.
Image Courtesy : Chip Somodevilla/Getty Images
AI was among the major issues on the agenda when Trump and Xi met at the White House on September 24. Xi used his remarks to call for AI development to remain under human control and said the United States and China have both the capability and responsibility to develop and manage AI in ways that serve the public. Trump, meanwhile, had said before the meeting that "super intelligence" would be discussed but that he did not intend to place new limits on AI development.
The contrasting approaches illustrate one of the central challenges surrounding the discussions. Both governments have an interest in maintaining their positions in the global AI race, but both also face concerns about increasingly powerful systems being used for cyberattacks, biological misuse, attacks on critical infrastructure and other security threats. U.S. and Chinese officials have therefore begun exploring whether there are specific risks on which the two countries could communicate even while competing aggressively in AI development.
The technology industry has become an important part of that conversation. Executives from major AI companies have been calling for greater international coordination on AI safety, arguing that the risks associated with increasingly capable systems may not stop at national borders. OpenAI and Anthropic leaders have both recently urged governments to establish international approaches to AI safety, adding pressure to discussions happening alongside the Trump-Xi meeting.
Those calls come as some leading AI researchers and executives have raised concerns about the possibility of future systems becoming increasingly difficult for humans to control. The debate is particularly focused on frontier AI models that can perform complex reasoning, operate computer systems, write software and potentially improve aspects of their own capabilities. The degree of risk associated with these systems remains contested, but concern about potential loss of control has become an increasingly prominent issue among researchers and technology leaders.
U.S.-China discussions about AI safety did not begin with the White House meeting. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng discussed the possibility of establishing an AI safety notification mechanism during meetings in New York earlier in September. The proposed mechanism would provide a channel for the two governments to notify one another about significant AI-related incidents and discuss shared risks.
Bessent subsequently described the proposed framework as a potential "U.S.-China AI Dialogue." The discussions could eventually cover issues including national security, transparency, significant AI incidents and ways to avoid a complete technological separation between the two countries. However, Chinese officials have not publicly committed to the full proposal, and significant disagreements remain between Washington and Beijing over technology restrictions and national security.
One possible outcome being discussed is an AI equivalent of the communication channels that have historically been used to reduce the risk of misunderstandings between rival powers. Analysts have suggested that a direct channel for communicating about serious AI incidents could become useful if an advanced system were to behave unexpectedly or if one country's AI activity were interpreted as a threat by the other.
Such a mechanism would not necessarily amount to a comprehensive global AI treaty. Instead, it could begin with narrower areas where both governments have an obvious interest in avoiding dangerous misunderstandings. Establishing reliable communication could be particularly important as AI systems become increasingly integrated into cybersecurity, military planning, financial systems and other critical infrastructure.
The involvement of private-sector AI companies adds another layer to the negotiations. American companies including OpenAI and Anthropic are developing some of the world's most advanced AI systems, while Chinese companies and research organizations are rapidly advancing their own models. Both countries therefore have access to technical expertise that governments may need when developing realistic safety standards.
At the same time, the technology companies have their own interests in the debate. U.S. AI companies want to maintain access to global markets and continue developing increasingly capable systems, while governments are concerned with national security, economic competitiveness and potential misuse. Those interests do not always align, meaning the presence of industry executives at discussions surrounding AI governance does not automatically translate into agreement over what rules should exist.
The disagreement is particularly visible in the United States. Some AI executives and researchers have advocated stronger safeguards or deliberately slowing certain aspects of frontier development, while the Trump administration has emphasized maintaining U.S. technological leadership. Trump said earlier in September that the United States already has tools to address AI-related misconduct and has repeatedly emphasized the importance of winning the competition with China.
China has taken a different approach to AI governance, having developed a broader regulatory framework covering areas such as generative AI, data and algorithmic systems. Chinese officials have also promoted the country's Global AI Governance Initiative as a framework for international cooperation. Analysts note that Beijing's approach combines AI development with government oversight and requirements that reflect China's political system.
That difference makes a comprehensive agreement between Washington and Beijing difficult. The two countries disagree over technology exports, advanced semiconductor access, cybersecurity and the broader strategic implications of AI. U.S. restrictions on high-end AI technology have also become a major point of tension, while American officials have accused Chinese companies of using techniques such as model distillation to obtain capabilities from U.S. systems. Chinese officials have rejected some of those accusations and criticized what they describe as attempts to use safety concerns to restrict China's technological development.
Nevertheless, AI creates an unusual area in which the two competitors share some common risks. A major cyberattack conducted with AI, an AI-related incident involving critical infrastructure or a serious failure involving an autonomous system could potentially affect multiple countries regardless of where the technology was developed. That makes communication and incident reporting potentially useful even when the countries remain competitors.
The question now is whether the discussions can produce something more concrete than diplomatic statements. Xi has publicly emphasized human control over AI, while Trump has indicated that he does not want to impose limits on the development of superintelligence. The two positions leave considerable room for disagreement over what "AI safety" should actually mean and how much government intervention should be involved.
For the technology industry, the outcome could have long-term implications. International AI standards could influence how companies develop models, test advanced systems, report safety incidents and deploy autonomous agents. They could also affect how AI companies operate across borders as governments establish different requirements for data, computing infrastructure and high-end AI models.
For now, the Trump-Xi discussions appear to be focused more on opening channels of communication and identifying shared risks than establishing a single global AI rulebook. The proposed U.S.-China notification mechanism and continuing AI dialogue could provide a foundation for future discussions, but the two countries remain divided over the balance between safety, regulation, national security and technological competition.
The significance of the meeting ultimately extends beyond the White House and Beijing. As AI becomes increasingly capable and increasingly important to national economies and security systems, decisions made by governments and technology companies could shape how the technology develops for years to come. Bringing AI executives into the conversation gives policymakers access to the people building these systems, while the involvement of national leaders underscores just how consequential the technology has become.
The Trump-Xi discussions therefore represent another step in an emerging global debate over who should establish the rules for advanced AI and how countries can manage shared risks while continuing to compete. Whether that dialogue develops into formal agreements remains unresolved, but the participation of both governments and the companies building frontier AI ensures that the issue is unlikely to disappear from the international agenda anytime soon.