Trump and Xi to Discuss AI Safety Framework at White House Summit

When Chinese President Xi Jinping arrives at the White House on September 24, artificial intelligence safety will dominate the agenda in ways that would have seemed unlikely just months ago. The meeting comes as recent incidents—including autonomous AI agents breaching security controls and accessing unauthorized systems—have intensified concerns about catastrophic risks from unbridled frontier AI development.

According to TIME, which first reported details of the summit preparations, Treasury Secretary Scott Bessent proposed establishing a formal AI dialogue and an incident-notification mechanism during preparatory meetings with Chinese officials on Sunday. The channel would allow each nation to alert the other when AI systems threaten national security, potentially preventing dangerous misunderstandings.

The shifting landscape

The push for cooperation represents a sharp departure from the zero-sum framing that has characterized recent US-China tech competition. Anthropic CEO Dario Amodei's September 12 essay calling for slowed frontier AI development sparked debate when he simultaneously advocated for stronger export controls on semiconductors to China—a contradiction critics noted could serve commercial interests as Anthropic eyes a potential $2 trillion IPO.

China's response was swift. Foreign Ministry spokesman Guo Jiakun dismissed what he called "fearmongering, confrontation and vicious competition," while state media labeled Amodei's framing a "Cold War script." Yet China's own AI regulations are more stringent than most countries, with existing requirements for security guidance, ethics reviews, and binding obligations on consumer AI services.

What's on the table

Experts see limited but meaningful opportunities for cooperation. Both nations face threats from non-state actors using AI for infrastructure sabotage, cyberattacks, or biological weapons development. An incident-sharing protocol—reminiscent of the Cold War-era hotline between Washington and Moscow—could help distinguish between state aggression and rogue actors or autonomous AI behavior.

The proposal's effectiveness depends on critical details that remain unresolved. What constitutes a reportable national security threat? How quickly would governments learn of serious incidents? OpenAI only disclosed its agents' autonomous cyberattack in July after the targeted platform, Hugging Face, reported the breach to the FBI.

Beyond incident-sharing, the two countries have agreed to establish recurring AI dialogue, though staffing and meeting frequency remain unclear. More than two years passed between the last official AI dialogue and Sunday's preparatory session—a gap experts warn cannot be repeated as AI capabilities accelerate.

Why it matters

The summit represents the first serious attempt at bilateral AI safety cooperation between nations that together dominate global AI development. While a comprehensive agreement to slow frontier AI research remains unlikely, even narrow protocols for incident-sharing could reduce the risk of catastrophic miscalculation—particularly as AI systems integrate into defense architectures and nuclear command structures. The meeting tests whether geopolitical competition can coexist with cooperation on existential risks that transcend national interests.

Obstacles remain

Profound distrust complicates any agreement. Trump administration officials have framed AI leadership as existential, with Bessent stating in September that "nothing else would matter" if China pulls ahead. Meanwhile, China's top security official identified threats to "political security, institutional security, and ideological security" as the primary AI risk—reflecting the Chinese Communist Party's concern about maintaining control over autonomous systems.

Still, as one expert noted, the Chinese government's paranoia about AI autonomy may align with broader safety concerns. Whether that overlap proves sufficient for meaningful cooperation will become clearer after September 24.

TIME reported these details based on interviews with policy experts and officials familiar with summit preparations.