US and China Weigh First Official AI Risk Talks Ahead of Trump-Xi Beijing Summit

--- headline: "US and China Weigh First Official AI Risk Talks Ahead of Trump-Xi Beijing Summit" slug: us-china-ai-talks-trump-xi-summit category: policy story_number: "14" date: 2026-05-08 sources: - name: The Wall Street Journal url: https://www.wsj.com/world/china/u-s-and-china-pursue-guardrails-to-stop-ai-rivalry-from-spiraling-into-crisis-4c50bd70 domain: wsj.com - name: Benzinga url: https://www.benzinga.com/markets/tech/26/05/52361259/trump-administration-eyes-first-official-ai-dialogue-with-china-at-beijing-summit-with-xi-jinping-report domain: benzinga.com - name: Digital Watch Observatory url: https://dig.watch/updates/us-china-ai-discussions-summit domain: dig.watch - name: Invezz url: https://invezz.com/news/2026/05/07/trump-xi-summit-trade-taiwan-ai-talks-in-focus-major-breakthroughs-unlikely/ domain: invezz.com - name: CommonWealth Magazine url: https://english.cw.com.tw/article/article.action?id=4760 domain: english.cw.com.tw - name: Fox Business url: https://www.foxbusiness.com/politics/white-house-china-industrial-scale-ai-technology-theft-trump-xi-summit domain: foxbusiness.com ---

The United States and China are quietly exploring the possibility of launching their first formal dialogue on artificial intelligence risks under the Trump administration, a move that would place the volatile question of AI safety at the center of President Donald Trump's upcoming summit with Chinese President Xi Jinping in Beijing on May 14-15.

The proposed talks, first reported by The Wall Street Journal and confirmed by multiple outlets, would focus on three core threat areas: unpredictable behavior from frontier AI models, the proliferation of autonomous military systems, and the potential for non-state actors to weaponize powerful open-source AI tools. If the two leaders agree to put AI on the formal summit agenda, it would mark the first direct AI-specific engagement between Washington and Beijing since Trump took office.

US Treasury Secretary Scott Bessent is leading the American side of the preliminary discussions, according to people familiar with the matter, while Chinese Vice Finance Minister Liao Min has been involved in early conversations about establishing the channel. Washington is still waiting for Beijing to formally designate a counterpart for the talks, a step that analysts say will signal how seriously China intends to engage.

Liu Pengyu, spokesperson for the Chinese Embassy in Washington, indicated that Beijing is open to the dialogue. He told reporters that China is ready to engage in communication on AI risk mitigation, a notably conciliatory statement given the heated rhetoric between the two governments on technology issues in recent weeks.

A Troubled Precedent

The proposed talks would build on -- and attempt to improve upon -- a prior round of AI diplomacy that produced limited results. In November 2023, then-President Joe Biden and Xi launched a formal AI dialogue at their California summit, with both sides eventually agreeing that humans, not AI, would retain authority over nuclear launch decisions. But the initiative stalled. Georgetown scholar Rush Doshi, who led the US side of those talks under Biden, told The Wall Street Journal that China placed its foreign ministry rather than technical experts in charge of negotiations, limiting the depth of engagement. "The US should have pushed harder for deeper technical engagement," Doshi said, a critique that frames the current effort as a second chance to get the structure right.

The 2024 agreement on human control over nuclear weapons represented a tangible, if narrow, accomplishment. But broader questions about AI safety testing, model transparency, and military applications remained unresolved. Non-governmental discussions have continued in parallel, with figures such as former Microsoft research executive Craig Mundie engaging Chinese counterparts from Tsinghua University and major Chinese AI companies on frontier-model safety and technical guardrails.

The Tension Underneath

Any AI safety dialogue will unfold against a backdrop of deepening distrust. Just weeks before the summit, the White House accused China-linked groups of conducting "industrial-scale" campaigns to steal US frontier AI models, alleging that tens of thousands of proxy accounts were used to exfiltrate proprietary technology from American labs. The Trump administration warned it would step up enforcement against the unauthorized distillation of advanced AI models, a technique in which smaller models are trained to replicate the capabilities of larger, proprietary systems.

The competitive gap between the two countries is also narrowing rapidly. Stanford University's 2026 AI Index found that the US performance advantage over China has nearly disappeared, with the leading American model ahead of the best Chinese model by just 2.7 percent on the Arena Leaderboard as of March 2026. That shrinking margin adds urgency to both the competitive and cooperative dimensions of the relationship.

In April, a bipartisan group in Congress introduced the Multilateral Alignment of Technology Controls on Hardware (MATCH) Act, which seeks to tighten semiconductor export controls by pressing US allies to close gaps that have allowed China to continue acquiring advanced chip-making equipment. Meanwhile, Xi has called for "original and disruptive innovation" in AI, signaling that China intends to compete on the frontier rather than simply follow.

What a Deal Could Look Like

Analysts who have tracked the bilateral AI relationship say the most realistic near-term outcome would be procedural rather than substantive: establishing a regular channel for communication rather than agreeing on binding rules. Some have raised the possibility of an AI hotline between senior leaders, modeled on Cold War-era nuclear communication channels, which could serve as a crisis-management tool if an AI system causes an unintended military escalation or a major civilian disruption.

The question of who leads the talks on the Chinese side will be a telling indicator. If Beijing assigns technical experts from its Ministry of Science and Technology or the Chinese Academy of Sciences rather than foreign ministry diplomats, it would suggest a willingness to move beyond symbolic gestures toward substantive engagement on model safety, testing protocols, and information-sharing about AI incidents.

Why This Matters

The proposed dialogue matters less for what it might produce at the summit itself -- most analysts expect no binding commitments -- than for the precedent it would set. Establishing a formal, ongoing AI risk channel between the world's two leading AI powers would acknowledge that advanced AI is a strategic stability issue, not merely an economic or technological race. In a year when both countries are pouring hundreds of billions of dollars into AI infrastructure and racing to build increasingly powerful models, even a modest framework for managing shared risks could reduce the probability of a catastrophic miscalculation.

What to Watch

The summit agenda is still being finalized, and whether AI makes the cut will depend on decisions by both Trump and Xi in the coming days. Watch for signals on three fronts: whether Beijing names a technical counterpart to Bessent, whether the final joint statement includes specific AI language, and whether any follow-up meetings are scheduled. If all three materialize, the May summit could mark the beginning of a genuine, if fragile, AI risk management architecture between the two nations most capable of building -- and most vulnerable to -- the technology's most dangerous applications.