Donald Trump and Xi Jinping are likely to discuss AI safety when they meet on Thursday, as officials from both countries weigh an emergency communication channel for AI-related security incidents. But export controls on advanced Nvidia chips and accusations of Chinese "distillation" remain unresolved, and neither side shows signs of slowing its AI development.
Ahead of the summit, U.S. Treasury Secretary Scott Bessent discussed the prospect of a U.S.-China AI dialogue with his Chinese counterpart, He Lifeng, over the weekend, including a channel for incidents up to a national-security level. The talks come as AI systems have taken more consequential actions online, including an OpenAI research model that gained unauthorized access to parts of the Hugging Face platform, and Anthropic's discovery that AI was being used to automate parts of cyberattacks.
"Leaders in both countries are spooked by the increasing cyber capabilities of AI models, and especially agents acting autonomously to escape containment and breach websites," Aalok Mehta, director of the Wadhwani AI Center at the Center for Strategic and International Studies, told CNBC.
Chip export controls remain a sticking point
Washington continues to restrict China's access to Nvidia's most advanced AI chips, even though several Chinese firms have reportedly been able to access the chips' compute power remotely through data centers in Southeast Asia. U.S. Trade Representative Jamieson Greer said chip controls were not on the agenda during preliminary talks over the weekend, but China could push to have them relaxed, according to Melanie Hart, senior director of the Atlantic Council's Global China Hub.
Chris McGuire, senior fellow for China and emerging technologies at the Council on Foreign Relations, said an agreement to slow AI's development pace is extremely unlikely. He noted that China has historically used AI dialogues to complain about U.S. export controls.
Distillation accusations add friction
Washington has also accused Chinese AI labs of "distillation" — training models on the output of more advanced U.S. systems — a claim Beijing rejects. In July, Bessent threatened sanctions against companies engaged in the practice, saying the administration had found watermarks of U.S. large language models on many of the Chinese models.
Trump has previously dismissed warnings that AI poses an existential threat, even as OpenAI's Sam Altman and Anthropic's Dario Amodei called for an industry slowdown. Yet in a Truth Social post on Monday, Trump said the government "will rein things in if we have to."
Mehta said progress on AI cooperation depends on two conditions: that it doesn't hurt either country's ability to compete on the merits of its technology, and that a verification mechanism exists so policymakers aren't relying solely on each other's word.
Source: CNBC
Trading involves risk.
Risk Disclaimer
