U.S. and China Seek AI Safety Dialogue Amid Tech Rivalry

Competing powers explore narrow path for AI cooperation

The United States and China are attempting to establish dialogue on artificial intelligence safety risks, even as the world's two largest economies remain locked in competition for AI dominance. U.S. Treasury Secretary Scott Bessent announced Sunday that discussions with Chinese Vice Premier He Lifeng included proposals for an AI dialogue mechanism and a notification system for AI incidents affecting national security, according to reporting by the Associated Press.

The talks preceded an expected meeting this week between President Donald Trump and Chinese leader Xi Jinping, with AI emerging as both a point of contention and potential cooperation between the rival nations.

Why it matters

As AI systems grow more powerful and autonomous, the risk of catastrophic failures, cyberattacks, or loss of control increases for both nations. Establishing communication channels for AI emergencies could prevent misunderstandings and cascading crises, even if broader technology competition continues. The narrow window for cooperation reflects how AI has become both a national security priority and a shared global risk.

Shared risks create opening for talks

Both governments face similar threats from advanced AI systems, including AI-enabled cyberattacks, biological misuse, major model failures, and potential loss of human control over autonomous systems. Sun Chenghao, a senior fellow at Tsinghua University's Center for International Security and Strategy, noted that China is paying growing attention to loss-of-control risks and cyber threats.

Michael Kratsios, Trump's science and technology adviser, told Fox News Sunday that avoiding shared risks is a key conversation topic with China. Chinese governance documents identify finance, electricity, telecommunications, and transportation as sectors potentially exposed to frontier AI risks.

Export controls remain firm despite Chinese advances

Any loosening of U.S. chip export restrictions appears off the table, according to Alfredo Montufar-Helu of consultancy Ankura. The U.S. has restricted China's access to advanced semiconductors to slow its AI development, though Chinese AI models from companies like DeepSeek and Moonshot AI have increasingly challenged American frontier models this year.

Professor Will Cong of Nanyang Technological University in Singapore noted that export controls have slowed China's expansion rate by limiting available computing power, but have not stopped development of competitive models. Chinese companies are increasingly turning to domestic chip manufacturers, giving Xi more confidence in negotiations.

Distillation accusations strain relations

U.S. allegations that Chinese AI labs are extracting capabilities from American models through a process called "distillation" have become a friction point. Anthropic has accused DeepSeek and Moonshot of routing user requests to its Claude AI models. Chinese officials reject these claims, arguing that distillation is common industry practice and not the primary driver of Chinese innovation.

Meanwhile, Beijing has expressed concerns about threats to China's critical infrastructure from American AI systems and the potential for AI to spread political misinformation.

Xiao Qian, vice dean of Tsinghua University's Institute for AI International Governance, warned that meaningful cooperation requires both sides to stop treating the other's technological progress as inherently threatening, though she acknowledged that disagreements make dialogue more necessary.

These details were first reported by the Associated Press.