On 24 September 2026, United States President Donald J. Trump hosted Chinese President Xi Jinping for a summit in Washington, DC. The two leaders discussed a range of issues, among which artificial intelligence (AI) featured prominently. Building on their May 2026 summit in Beijing, this meeting prioritized bilateral talks, trade and industry representation, a hotline for reporting AI incidents and a deadline of November 2026 for a follow-on technical dialogue. This essay explores these shifts and what is needed to build on the achievements of the September summit at the follow-on dialogue and beyond.
Evolving bilateral engagement on AI
The September 2026 summit between Trump and Xi marked a significant departure from past bilateral interactions regarding AI between China and the USA. A first notable difference was in the format of the meeting. As US president, Joe Biden held two summits with Xi where AI-related issues were discussed, but both took place on the sidelines of Asia-Pacific Economic Cooperation (APEC) meetings: the first in San Francisco in November 2023, the second in Lima, Peru, a year later.
By contrast, Trump travelled to Beijing in May 2026 for a dedicated bilateral summit with Xi that generated economic and security agreements. Similarly, the September 2026 summit was the main purpose of Xi’s state visit to the USA. Xi went straight to Washington, DC, despite the fact that a United Nations General Assembly session was taking place in New York. Given that China has made a point of its support for multilateralism, and for the UN in particular, this decision sent a strong message about the value Xi placed on holding direct bilateral talks, especially when it came to trade and AI safety.
Another significant shift was in participation. At both the May and September summits, the representation of both Chinese and US trade officialsas well as industry representatives was stronger than in previous Xi–Biden summits, allowing the summit to address more concrete AI-related mechanisms. The November 2023 and 2024 summits were dominated by Chinese Ministry of Foreign Affairs and US Department of State officials, who are chiefly responsible for political relations. Only a handful of trade officials were present, and no reports of the meetings indicate that industry was represented.
By contrast, at the May 2026 summit, chief executive officers (CEOs) from Chinese technology companies such as Bytedance, Lenovo and Xiaomi were invited to join the state banquet, alongside China’s Vice Premier He Lifeng. US Treasury Secretary Scott Bessent and Trade Representative Jamieson Greer led a delegation that included the CEOs of US firms working on AI-related technologies, including Apple, Boeing, Coherent, GE Aerospace, Illumina, Meta, Micron Technology, Nvidia, Qualcomm and SpaceX.
The September 2026 summit similarly featured the US Treasury Secretary and Trade Representative, as well as Vice Premier He Lifeng and China’s International Trade Representative, Li Chenggang. It included a state dinner that featured over 20 CEOs, including from such AI industry players as Advanced Micro Devices, Apple, Dell, GE Aerospace, Google, Lockheed Martin, Meta, Micron Technology, Microsoft, Nvidia, OpenAI, SpaceX and Zoom.
A final major differencewas in the scope and outcomes. The November 2023 and 2024 meetings created a bilateral AI consultative channel at intergovernmental level and a political statement on ‘the need to maintain human control over the decision to use nuclear weapons’, but discussions skewed towards normative pledges rather than technical substance.
By contrast, the May 2026 meeting covered a range of substantive trade issues, including AI, though not more intractable issues like access to semiconductors by China and critical minerals by the USA. Nevertheless, it laid the groundwork for targeted AI talks at the Washington summit four months later.
The September 2026 summit yielded a deadline of November for a follow-on dialogue to facilitate exchanges on AI-related risks and opportunities, as well as the establishment of a bilateral communication channel for AI incidents, dubbed by some an ‘AI hotline’. Such a mechanism is integral to future risk reduction and is likely to be a centrepiece of future bilateral dialogue.
The potential for achieving concrete progress on AI in upcoming bilateral China–USA talks appears to be stronger than it has been for some time. However, along with the opportunities, there are also some issues that could prove challenging.
Agreeing on terminology
The US official summary of the September 2026 summit indicates that both sides agreed to use ‘super intelligence’ or ‘超级智能’ to describe the suite of ‘applicable emerging technologies’ under discussion. However, the official report from China continues to use ‘artificial intelligence’ or ‘人工智能’. This may seem like mere semantics, but it is integral to how China and the USA frame not only their talks, but also their mechanisms.
The term ‘super intelligence’has long been used to denote any hypothetical intellect that far surpasses human cognitive capabilities in almost all domains. It is most associated with future developments in machine intelligence, with systems developing this superintelligence through recursive self-improvement. This stands in contrast to traditional definitions of ‘artificial intelligence’, as mimicking human intelligence and only surpassing human capabilities in specific and limited ways.
While the USA’s focus on what it terms ‘super intelligence’ indicates the potential direction of AI development and the risks associated with losing human control, it also suggests a forward-looking approach that is based on hypothetical trends. China’s decision to retain the term ‘artificial intelligence’ suggests a focus on current conditions and may create a longer-term tension with the US preoccupation with future trends.
Fleshing out the AI hotline proposal
Another open question that the upcoming dialogue needs to address is how the AI hotline is to be designed and used: what kinds of incident would be reported to the other state using the hotline, and what response procedures would apply. The above terminological divergence arguably adds to this lack of clarity over what to report, particularly with novel incidents.
Moreover, there remains an issue over how reporting would happen and whether the responsibility should be delegated to the agencies closer to the detection and response processes. Anomalous behaviours by AI models are often first detected by the companies developing and operating them. However, it remains unclear whether labs working on frontier AI models would be required to report such incidents through the AI hotline, and how they should do so.
In China, there already exists a tiered reporting system under the Cyberspace Administration of China. In the USA, however, there is no single, mandatory national reporting system that covers all general anomalous AI incidents. Instead, the US channels for reporting often depend on the nature of the incident, the infrastructure affected, agency-specific rules and private industry frameworks.
Clarifying these domestic reporting processes is essential for determining both the types of incident that should be reported and the points of contact for any bilateral mechanism between China and the USA.
Integrating trade and industry expertise
The involvement of industry and technical experts is essential to the success of ongoing engagement on AI. This means that there is a pressing need to move industry involvement from the banquet table to the dialogue table. Although companies may be disinclined to limit their own AI development, they can help to translate normative commitments into concrete measures and procedures that are acceptable to and feasible for both sides’ AI industries. This is particularly relevant when it comes to delineating roles and reporting mechanisms within industry and trade departments, which will be crucial as both sides decide on their designated points of contact and the types of incident data to be shared between them.
Both the Chinese and US official summaries of the September 2026 summit indicate that neither side addressed contentious AI-related trade issues that dominated earlier summits, such as US limits on China’s access to semiconductors, advanced chips and computing power, and Chinese limits on exports of critical minerals that are essential in precision chip-making machinery, datacentre cooling systems and physical robotics. Given that both sides control something the other needs for AI development—in terms of technology and resources—this symmetry of interests could serve as a launchpad for more specifics coming out of the upcoming dialogue.
Building on shifts that began in May 2026, the September summit made welcome progress on the format, participation, scope and outcomes of AI-related talks between China and the USA. Whether the upcoming dialogue builds further on these to yield definitions of AI-related incidents, designated points of contact and symmetric information sharingwill serve as a key indicator of the future trajectory of bilateral talks and action on AI issues.
Most importantly, if the upcoming dialogue features technical and operational inputs from both trade and industry—which drive AI trends and integration—it has a better chance of producing other practical measures and mechanisms to be agreed on in future China–USA presidential summits and technical dialogues.
This essay was made possible by a grant from the Andrew Carnegie Foundation, formerly Carnegie Corporation of New York.
ABOUT THE AUTHOR(S)
Dr Lora Saalman is an Associate Senior Fellow within SIPRI’s Armament and Disarmament and Conflict, Peace and Security research areas.
Fei Su is a Senior Researcher in the SIPRI China and Asia Security Programme.
On 24 September 2026, United States President Donald J. Trump hosted Chinese President Xi Jinping for a summit in Washington, DC. The two leaders discussed a range of issues, among which artificial intelligence (AI) featured prominently. Building on their May 2026 summit in Beijing, this meeting prioritized bilateral talks, trade and industry representation, a hotline for reporting AI incidents and a deadline of November 2026 for a follow-on technical dialogue. This essay explores these shifts and what is needed to build on the achievements of the September summit at the follow-on dialogue and beyond.
Evolving bilateral engagement on AI
The September 2026 summit between Trump and Xi marked a significant departure from past bilateral interactions regarding AI between China and the USA. A first notable difference was in the format of the meeting. As US president, Joe Biden held two summits with Xi where AI-related issues were discussed, but both took place on the sidelines of Asia-Pacific Economic Cooperation (APEC) meetings: the first in San Francisco in November 2023, the second in Lima, Peru, a year later.
By contrast, Trump travelled to Beijing in May 2026 for a dedicated bilateral summit with Xi that generated economic and security agreements. Similarly, the September 2026 summit was the main purpose of Xi’s state visit to the USA. Xi went straight to Washington, DC, despite the fact that a United Nations General Assembly session was taking place in New York. Given that China has made a point of its support for multilateralism, and for the UN in particular, this decision sent a strong message about the value Xi placed on holding direct bilateral talks, especially when it came to trade and AI safety.
Another significant shift was in participation. At both the May and September summits, the representation of both Chinese and US trade officials as well as industry representatives was stronger than in previous Xi–Biden summits, allowing the summit to address more concrete AI-related mechanisms. The November 2023 and 2024 summits were dominated by Chinese Ministry of Foreign Affairs and US Department of State officials, who are chiefly responsible for political relations. Only a handful of trade officials were present, and no reports of the meetings indicate that industry was represented.
By contrast, at the May 2026 summit, chief executive officers (CEOs) from Chinese technology companies such as Bytedance, Lenovo and Xiaomi were invited to join the state banquet, alongside China’s Vice Premier He Lifeng. US Treasury Secretary Scott Bessent and Trade Representative Jamieson Greer led a delegation that included the CEOs of US firms working on AI-related technologies, including Apple, Boeing, Coherent, GE Aerospace, Illumina, Meta, Micron Technology, Nvidia, Qualcomm and SpaceX.
The September 2026 summit similarly featured the US Treasury Secretary and Trade Representative, as well as Vice Premier He Lifeng and China’s International Trade Representative, Li Chenggang. It included a state dinner that featured over 20 CEOs, including from such AI industry players as Advanced Micro Devices, Apple, Dell, GE Aerospace, Google, Lockheed Martin, Meta, Micron Technology, Microsoft, Nvidia, OpenAI, SpaceX and Zoom.
A final major difference was in the scope and outcomes. The November 2023 and 2024 meetings created a bilateral AI consultative channel at intergovernmental level and a political statement on ‘the need to maintain human control over the decision to use nuclear weapons’, but discussions skewed towards normative pledges rather than technical substance.
By contrast, the May 2026 meeting covered a range of substantive trade issues, including AI, though not more intractable issues like access to semiconductors by China and critical minerals by the USA. Nevertheless, it laid the groundwork for targeted AI talks at the Washington summit four months later.
The September 2026 summit yielded a deadline of November for a follow-on dialogue to facilitate exchanges on AI-related risks and opportunities, as well as the establishment of a bilateral communication channel for AI incidents, dubbed by some an ‘AI hotline’. Such a mechanism is integral to future risk reduction and is likely to be a centrepiece of future bilateral dialogue.
The potential for achieving concrete progress on AI in upcoming bilateral China–USA talks appears to be stronger than it has been for some time. However, along with the opportunities, there are also some issues that could prove challenging.
Agreeing on terminology
The US official summary of the September 2026 summit indicates that both sides agreed to use ‘super intelligence’ or ‘超级智能’ to describe the suite of ‘applicable emerging technologies’ under discussion. However, the official report from China continues to use ‘artificial intelligence’ or ‘人工智能’. This may seem like mere semantics, but it is integral to how China and the USA frame not only their talks, but also their mechanisms.
The term ‘super intelligence’ has long been used to denote any hypothetical intellect that far surpasses human cognitive capabilities in almost all domains. It is most associated with future developments in machine intelligence, with systems developing this superintelligence through recursive self-improvement. This stands in contrast to traditional definitions of ‘artificial intelligence’, as mimicking human intelligence and only surpassing human capabilities in specific and limited ways.
While the USA’s focus on what it terms ‘super intelligence’ indicates the potential direction of AI development and the risks associated with losing human control, it also suggests a forward-looking approach that is based on hypothetical trends. China’s decision to retain the term ‘artificial intelligence’ suggests a focus on current conditions and may create a longer-term tension with the US preoccupation with future trends.
Fleshing out the AI hotline proposal
Another open question that the upcoming dialogue needs to address is how the AI hotline is to be designed and used: what kinds of incident would be reported to the other state using the hotline, and what response procedures would apply. The above terminological divergence arguably adds to this lack of clarity over what to report, particularly with novel incidents.
Moreover, there remains an issue over how reporting would happen and whether the responsibility should be delegated to the agencies closer to the detection and response processes. Anomalous behaviours by AI models are often first detected by the companies developing and operating them. However, it remains unclear whether labs working on frontier AI models would be required to report such incidents through the AI hotline, and how they should do so.
In China, there already exists a tiered reporting system under the Cyberspace Administration of China. In the USA, however, there is no single, mandatory national reporting system that covers all general anomalous AI incidents. Instead, the US channels for reporting often depend on the nature of the incident, the infrastructure affected, agency-specific rules and private industry frameworks.
Clarifying these domestic reporting processes is essential for determining both the types of incident that should be reported and the points of contact for any bilateral mechanism between China and the USA.
Integrating trade and industry expertise
The involvement of industry and technical experts is essential to the success of ongoing engagement on AI. This means that there is a pressing need to move industry involvement from the banquet table to the dialogue table. Although companies may be disinclined to limit their own AI development, they can help to translate normative commitments into concrete measures and procedures that are acceptable to and feasible for both sides’ AI industries. This is particularly relevant when it comes to delineating roles and reporting mechanisms within industry and trade departments, which will be crucial as both sides decide on their designated points of contact and the types of incident data to be shared between them.
Both the Chinese and US official summaries of the September 2026 summit indicate that neither side addressed contentious AI-related trade issues that dominated earlier summits, such as US limits on China’s access to semiconductors, advanced chips and computing power, and Chinese limits on exports of critical minerals that are essential in precision chip-making machinery, datacentre cooling systems and physical robotics. Given that both sides control something the other needs for AI development—in terms of technology and resources—this symmetry of interests could serve as a launchpad for more specifics coming out of the upcoming dialogue.
Building on shifts that began in May 2026, the September summit made welcome progress on the format, participation, scope and outcomes of AI-related talks between China and the USA. Whether the upcoming dialogue builds further on these to yield definitions of AI-related incidents, designated points of contact and symmetric information sharingwill serve as a key indicator of the future trajectory of bilateral talks and action on AI issues.
Most importantly, if the upcoming dialogue features technical and operational inputs from both trade and industry—which drive AI trends and integration—it has a better chance of producing other practical measures and mechanisms to be agreed on in future China–USA presidential summits and technical dialogues.
This essay was made possible by a grant from the Andrew Carnegie Foundation, formerly Carnegie Corporation of New York.
ABOUT THE AUTHOR(S)