US and China Seek Common Ground on AI Safety Amid Fierce Tech Rivalry

US President Donald Trump and Chinese leader Xi Jinping are expected to meet this week, with artificial intelligence safety emerging as a potential area of discussion between the two nations competing for AI dominance.

Ahead of the meeting, US Treasury Secretary Scott Bessent said after talks in New York with Chinese Vice Premier He Lifeng that the two countries had discussed setting up a mechanism for AI dialogue. Washington also proposed a new "notification mechanism" for AI incidents that could affect national security, Bessent told reporters. "We want a shared vision of common goals and common threats," he said.

Michael Kratsios, Trump's science and technology adviser, said it was important to hold conversations with China on AI. "The key thing that we want to talk to them about is avoiding shared risks, and that's something that I think we can have a fruitful conversation on," he said.

Analysts suggest the agenda could focus on risks that would harm both countries, including AI-enabled cyberattacks, biological misuse, major model failures and loss of human control. Sun Chenghao, a senior fellow at Tsinghua University's Center for International Security and Strategy, said there remains "meaningful room for cooperation" on such issues. "It would be a mistake to dismiss the entire safety debate as geopolitical competition," Sun said, noting that China itself is paying growing attention to loss-of-control risks, malicious use, cyber threats and the safety of AI agents.

Both governments are concerned about how AI could be used to attack critical infrastructure. Chinese governance documents identify finance, electricity, telecommunications and transportation as areas potentially exposed to frontier-AI risks. Samm Sacks, a senior fellow at the Johns Hopkins School of Advanced International Studies, said the Chinese side has shown real interest in discussing AI and cyber incident information sharing with US experts. Xiao Qian, vice dean of Tsinghua University's Institute for AI International Governance, said that if the two countries can discuss how to report and communicate during an AI-related emergency, "it can help the two countries to build technical confidence over time."

However, hopes for finding common ground are clouded by competition. As China narrows the technology gap with the US, each country has expressed concern over how AI in the other's hands could hurt them. Scott Singer, co-director of the China AI Initiative at the Carnegie Endowment for International Peace, said the US-China conversation on AI "is getting caught in the ups and downs in the broader US-China relationship."

Trump has repeatedly said "whoever wins AI, wins," insisting the US needs to keep a technological edge over China and downplaying calls by American tech leaders to slow AI's development over safety concerns. Beijing has accused the US of pursuing an AI monopoly as it curbs China's AI growth, including by restricting access to the world's most advanced chips. Any loosening of US export controls is likely to be off the table at the meeting, said Alfredo Montufar-Helu of the consultancy Ankura, who added that the room for collaboration is "quite narrow."

Despite restrictions, Chinese AI models are increasingly challenging American frontier models. Models led by Moonshot AI and DeepSeek this year have shown they can, by some measures, be as powerful as those of US rivals like Anthropic and OpenAI, often at lower cost, and many are making inroads globally, including in the US. Professor Will Cong of Nanyang Technological University in Singapore said export controls have not stopped China from developing frontier-competitive models but slow its rate of expansion by limiting the compute it can use to serve users. George Chen of The Asia Group said China's chip developers have been catching up with the US, giving Xi more confidence in his talks with Trump.

US claims that China is misappropriating American AI technology remain another sore point. The US government and AI companies have stepped up allegations of "malicious" extraction of capabilities, or "distillation," from top American models, claims Chinese officials have rejected. Anthropic has accused Chinese AI models from DeepSeek and Moonshot of routing user requests to its Claude models. Rebecca Arcesati of the Mercator Institute for China Studies said Chinese labs are likely attempting to "piggyback on the massive training and inference infrastructure of their US rivals" given US restrictions on China's compute power. Chinese officials and some technology analysts argue that distillation is not uncommon across the AI industry and is not the main explanation behind Chinese models' improving capabilities.

Beijing has also expressed worries over threats to China's critical infrastructure from American AI, including Anthropic's powerful Claude Mythos, as well as the risk of AI aiding the spread of political misinformation and potentially challenging its governance. "We cannot expect meaningful cooperation on AI safety if one side increasingly treats the other side's technological progress as a national security threat," Xiao said, adding that disagreements between the two countries "make dialogue even more necessary."