Expert SpeakDigital Frontiers
Published on Sep 30, 2026
As AI moves to the centre of US-China relations, Washington and Beijing are seeking to manage shared risks without easing their technology rivalry, while states wanting a say in the AI order must act before its terms are set
This article is part of the series, “Red Carpet, Red Lines: Xi–Trump and China’s Near-Peer Moment”
The three-day summit between Chinese President Xi Jinping and US President Donald Trump in Washington from 23 to 25 September 2026 <a href="https://www.aljazeera.com/news/2026/9/21/whats-the-us-china-ai-hotline-that-trump-plans-to-pitch-to-xi” rel=”nofollow noopener” target=”_blank”>was expected to cover tariffs, rare earths, the Iran war, sanctions on buyers of Russian energy and Taiwan. Yet the most consequential outcome appears to be on artificial intelligence (AI), a domain that was barely on the diplomatic radar a month ago.
AI has moved from the edge of US-China engagement to its centre. That shift reflects how both countries view technology as both a set up a communication channel for AI incidents shows that Washington and Beijing are learning to compete and cooperate at the same time
AI has moved from the edge of US-China engagement to its centre. That shift reflects how both countries view technology as both a
The spade work happened in New York on 20 September, three days before Xi’s arrival, where US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng held preparatory talks. The two sides reportedly agreed to establish a formal dialogue on AI and meet again in about two months, with Bessent naming Shenzhen as the probable venue. The summit formalised preparatory discussions on AI and cyber risks, establishing a US–China “Super Intelligence (SI) Dialogue” and a bilateral channel for AI incidents. China confirmed both, although its readout continued to refer to “AI” rather than “super intelligence”.
The Contours of AI Agenda
The AI track of the summit turned on four themes, which the outcomes addressed unevenly. First, the summit delivered the bilateral communication channel for AI incidents that Bessent had proposed: a US-China AI safety and national security notification mechanism that could alert both countries in real time to any AI-related national security concern. Existing US-China hotlines have had limited success; a military link set up in 2008, for example, has at times gone unanswered. The earlier AI dialogue broke down when frontier risks became entangled with other disputes. The new mechanism faces the same danger. As the trade truce has now been extended to 10 January 2027, the next round of talks is likely to take place under the shadow of a trade deadline, precisely the kind of linkage that has undermined previous efforts.
The most difficult question, which the fact sheet leaves open, is what counts as a trigger, particularly because attribution can be challenging in an AI-related crisis – determining whether a disruptive event was directed by a state, carried out by a non-state actor, or caused by an AI system operating beyond its intended limits. In a relationship already marked by deep mistrust, that kind of uncertainty could easily escalate a crisis.
The second risk comes from agentic and frontier systems, including incidents in which AI agents, unable to solve a problem during testing, break out of their sandbox and breach an external organisation’s infrastructure. Any such accident could involve an American system infiltrating Chinese critical infrastructure, or vice versa. Cyber-offensive capability and the misuse of AI’s biological knowledge fall into the same category. The summit did not, however, produce any agreement on jointly developing or regulating frontier models.
The third is the military dimension. In their 2024 pledge, former US President Joe Biden and Xi made a commitment to keep humans in control of decisions on nuclear use. Reaffirming or expanding this principle was one possible deliverable, but the fact sheet does not mention it, nor the growing integration of AI into intelligence, battle management and drone-swarm coordination.
The fourth is that US officials have made clear that export controls on advanced chips and semiconductor manufacturing equipment are not part of the new dialogue. The exclusion is deliberate: Washington wants a risk-reduction channel that leaves its technological leverage untouched.
Competition Beneath the Cooperation
Reading the hotline proposal as a thaw in the technology rivalry would be a mistake, as the competition between the two has sharpened. <a href="https://www.bcg.com/publications/2026/us-and-china-ai-<a href="https://absafricatv.com/7-days-countdown-to-the-caf-strategy-conference-in-cabo-verde/” title=”7 Days COUNTDOWN to the CAF Strategy Conference in Cabo Verde”>strategy-causing-global-ai-divide” rel=”nofollow noopener” target=”_blank”>Capital expenditure by leading US technology firms exceeded US$ 400 billion in 2025, compared with US$ 63 billion in China. The United States also retains a clear lead in frontier models, talent and data-centre capacity.
China, however, has pursued a different strategy, focusing on cost-optimised, largely open-weight models and rapid diffusion across its economy under the “AI+” initiative. It is also building a domestic semiconductor supply chain, and state-funded data centres must now use only indigenous chips.
A few days before the summit, for instance, Huawei announced a system that it claims can link up to a million processors into a single computer. China also holds structural advantages in rare earths and electricity generation, both critical inputs for the AI economy.
Washington, for its part, has shown that security will take precedence over the global spread of AI. In June 2026, a US export control directive forced Anthropic to suspend access to its Claude Mythos model. The overall picture is not a race with a clear finish line but two parallel ecosystems advancing side by side.
Implications for India and the Global South
For India, the summit’s AI track offers two lessons. First, frontier-risk governance is narrowing into a conversation largely between the two countries that build frontier models. India must work to ensure that any bilateral crisis mechanism complements, rather than replaces, inclusive platforms. These include the United Nations (UN) Independent International Scientific Panel on AI and the Global Dialogue on AI Governance.
Second, the bifurcation of the two technology stacks is gathering pace. India has so far avoided a binary choice by working with the United States on infrastructure, the European Union on governance and the United Arab Emirates (UAE) on compute, while engaging China through platforms such as BRICS. India also benefits from being the second-largest market for leading US AI firms. The cost of this multi-alignment, however, will rise as the two ecosystems become more self-reinforcing. New Delhi should use its scale while it still can, which means pushing for interoperability standards, building resilient domestic compute and securing a voice in any future incident-reporting frameworks that affect its critical infrastructure.
Conclusion
The summit’s own messaging on AI was more ambiguous. Trump had said “superintelligence” would be a major topic, but he also pointed to the US Department of Justice as America’s guardrail. Earlier, at the UN, he had rejected any “globalist” scheme for governing the technology. The message is that Washington wants crisis management, not regulation, and certainly not multilateral constraint.
Xi’s framing, in contrast, was more detailed. He said that the two countries should jointly guard against the misuse of AI rather than against each other, and that its development must remain under human control. He described the competition as one that should stay within bounds: a race to catch up rather than a contest with a single winner.
For other countries, the lesson is that the AI order is being shaped through bilateral bargaining between Washington and Beijing, and that states seeking a say must act before the terms are set.
The gap between the two framings matters, as Beijing presents itself as a responsible steward of AI and a champion of multilateral governance, while Washington keeps its distance from both. Nevertheless, a Chinese government spokesperson dismissed the call to pace frontier development as proposed by Anthropic CEO Dario Amodei.
Whether the AI dialogue becomes a durable institution or another short-lived gesture will be decided at the next round, due by November, and in the run-up to the January deadline.For other countries, the lesson is that the AI order is being shaped through bilateral bargaining between Washington and Beijing, and that states seeking a say must act before the terms are set.
Soumya Awasthiis a Fellow with the Centre for Security, Strategy and Technology at the Observer Research Foundation.
The views expressed above belong to the author(s). ORF research and analyses now available on Telegram! Click here to access our curated content — blogs, longforms and interviews.
Author
Soumya Awasthi
Dr Soumya Awasthi is a Fellow, Centre for Security, Strategy and Technology at the Observer Research Foundation.
Her work focuses on the intersection of technology and …
Related Search Terms
Publications
Xi and Trump: Resetting the Strategic Framework of US-China Relations?
China Foreign Policy
Sep 30, 2026
AI, Technology and the Evolving US-China Relationship
Artificial Intelligence
Sep 30, 2026
Series
