U.S.-China AI Safety Talks Face Trust Deficit Ahead of Summit

Summit to address AI governance amid mutual suspicion

Artificial intelligence governance will feature on the agenda when President Donald Trump meets Chinese leader Xi Jinping in Washington next week, according to a report first published by PBS NewsHour. The meeting comes as both superpowers pursue AI dominance while expressing deep mistrust of each other's intentions.

The challenge is stark: global AI safety strategy requires cooperation between the world's two leading AI powers, yet each views the other as the primary obstacle to responsible development. Trump has repeatedly warned that regulating AI would hand China an advantage, declaring that the nation leading on the technology "wins." Meanwhile, Chinese officials accuse the U.S. of attempting to monopolize the AI industry through suppression tactics.

Why it matters

Without some form of coordination between Washington and Beijing, efforts to manage existential risks from advanced AI systems will remain fragmented and potentially ineffective. The meeting represents a rare opportunity to establish even minimal dialogue on safety protocols, though experts expect any agreements to be largely symbolic rather than substantive.

Divergent development paths create friction

The two nations are advancing AI through fundamentally different strategies. U.S. companies leverage cutting-edge chips and expanding data center infrastructure to push frontier capabilities, while keeping most models closed-source. China, restricted from accessing the world's most advanced AI chips through U.S.-led export controls, has focused on developing open-source models designed for wide global adoption at lower cost.

Chinese AI models have nonetheless made significant advances and gained traction with American companies despite the chip restrictions. U.S. Treasury Secretary Scott Bessent said Tuesday that America needs to develop more open-source models to counter China's approach.

David Sacks, a top AI adviser to Trump, argued on Fox News that China has "already rejected" slowing development of advanced AI systems. Unlike Cold War nuclear verification regimes, he said, AI commitments would be impossible to verify. "In this case, we don't have trust and we can't verify," Sacks said. "So let's not be foolish about this and just hand the whole ball game to China."

Industry leaders call for dual approach

Leading American AI companies Anthropic and OpenAI have both restricted access in China while calling for continued U.S. technological leadership. Yet they've also advocated for cooperation on safety measures.

Anthropic CEO Dario Amodei wrote in a recent essay that "global pacing will require cooperation with China, the autocratic country with by far the most advanced AI capabilities." He simultaneously called for maintaining chip export restrictions to slow China's progress. His remarks drew sharp criticism from Beijing, with China's Ministry of Foreign Affairs calling them fearmongering and the state-run Global Times labeling them a "Cold War playbook."

Competing alliances emerge

Both nations have moved to build AI governance coalitions with aligned countries. China launched the World Artificial Intelligence Cooperation Organization in July with 29 founding members including Russia and Pakistan. The U.S. leads the Pax Silica initiative, focusing on collaboration with Japan, the U.K., and Australia on AI supply chains.

Samm Sacks, a senior fellow at Johns Hopkins School of Advanced International Studies, suggested expectations should remain modest. "A formal agreement or consensus may be near impossible, but they do not need to go that far," Sacks said. "Trump and Xi just need to create political space by acknowledging AI poses risks to both countries."

Details of the upcoming summit discussions were first reported by PBS NewsHour, with reporting by Chan Ho-him, Josh Boak, and Ken Moritsugu.