As President Donald Trump and Chinese President Xi Jinping convene in Beijing for their highly anticipated summit, the agenda is dominated by familiar issues: trade tariffs, Taiwan, and semiconductor export controls. However, a new and urgent topic has emerged as a focus of the bilateral discussions: the question of whether the two superpowers can agree on basic “guardrails” for artificial intelligence.
According to a report by AFP published in the Taipei Times, analysts and diplomats are cautiously optimistic that the summit could produce at least a rhetorical commitment to AI safety cooperation, even as they acknowledge the deep structural obstacles to any binding agreement.
The push for AI guardrails reflects a rare area of mutual concern between the two nations. Despite their intense technological rivalry, both Washington and Beijing recognize that the unchecked development of autonomous AI systems, particularly in the military domain, poses risks that neither side can fully manage alone.
(Related: Military AI Use Could Be on the Agenda When Trump and Xi Meet in Beijing)
A Shared Concern, Not a Shared Agenda
Michael Jinghan Zeng, a professor at City University of Hong Kong, described the dynamic as one of genuine but fragile convergence. “There is a kind of shared concern about where this AI arms race might be going,” Zeng told AFP, warning that it could create an “out of control” scenario. He noted that “despite critical disagreements on a wide range of issues, there is also this kind of understanding from both sides” on the need for some form of guardrails.
Zeng suggested that Xi and Trump could “commit to some rhetorical signal” in Beijing as a basis for further cooperation, a modest but meaningful step given the current state of the relationship.
The precedent for such a commitment exists. In 2024, President Xi agreed with then-President Biden that humans must remain in control of the decision to fire nuclear weapons, a principle that both sides have maintained as a baseline for AI safety discussions. Extending that principle to a broader range of autonomous military systems is now on the table.
The Biological Weapons Risk
Beyond the nuclear domain, experts are particularly focused on the risk of AI being used to develop biological weapons. Chris McGuire, writing for the Council on Foreign Relations, argued that this risk creates a powerful incentive for cooperation that transcends the bilateral rivalry. “If a non-state actor uses an AI model to develop a biological weapon, that could pose catastrophic risks to both the United States and China,” McGuire wrote. “Over the long term, addressing these risks will require cooperation.”
However, McGuire also cautioned that China’s “willingness to make and abide by robust international commitments on AI safety is low,” a sobering assessment that tempers expectations for what the summit can realistically achieve.
The Competition That Overshadows Cooperation
Chen Liang, founder of Strategic Times Consulting, offered a similarly measured view. He did not expect any “dramatic breakthroughs” from the summit, noting that competition is likely to remain stiff “in high-tech sectors like AI chips that directly involve the core interests of both sides.” Trump’s visit would merit attention, Chen suggested, if he and Xi manage to “shelve the most sensitive issues” while establishing “rule-based tracks” on points of cooperation.
Grace Shao, a China AI analyst and author of the AI Proem newsletter, framed the challenge in structural terms. “The talent, capital and supply chains underpinning the field are deeply intertwined across the United States and China,” she told AFP. “Any delusion of full decoupling isn’t realistic on any near-term horizon.” Shao argued that “leadership in the technology will define the next decade of productivity and growth, so it’s in everyone’s interest that the two superpowers find common ground on sensible guardrails for AI.”
The Contradiction at the Heart of the Talks
The push for AI guardrails at the summit exists in uncomfortable tension with other elements of U.S. policy. As we reported earlier today, the White House was alarmed by Chinese think tanks requesting access to Anthropic’s Mythos model, a model whose cybersecurity capabilities are so powerful that access has been restricted to approximately 40 partner organizations. Washington is simultaneously asking Beijing to cooperate on AI safety while actively working to prevent Chinese researchers from evaluating the most capable Western models.
This contradiction is not lost on Chinese officials. Beijing has refuted accusations of large-scale AI “distillation” of U.S. rivals and has pushed back against what it characterizes as an attempt by Washington to use AI safety frameworks as a cover for technological containment.
The summit’s executive delegation itself illustrates the complexity of the moment. Jensen Huang of Nvidia, Tim Cook of Apple, and Elon Musk of Tesla are all traveling with President Trump, a signal that commercial interests remain central to the bilateral relationship even as the two governments spar over AI governance. Whether the summit produces a meaningful agreement on guardrails or merely a vague statement of principles will depend on whether both sides can find enough common ground to overcome their competing interests.