Ex-US treasury secretary urges US, China to cooperate on AI de-risking

The United States (US) and China should draw lessons from the Cold War-era arms control mechanism to address the risks posed by artificial intelligence (AI), former US Treasury Secretary Robert Rubin said on Monday, urging the two countries to work together despite mutual distrust.

Speaking via video at the 2026 Bund Summit in Shanghai, Rubin said Washington and Beijing needed to set aside their antagonism and cooperate on AI safety, warning that both countries could pay a heavy price if they failed to tackle the risks.

The summit, themed "Technology, Geopolitics and the Global Economy," is jointly hosted by the China Finance 40 Forum and Tsinghua University from October 19 to 22.

Rubin, who served as US Treasury secretary from 1995 to 1999, made the remarks during a panel discussion with Huang Yiping, dean of Peking University's National School of Development.

"I think we've got to get past that to deal with these AI safety issues, or both of our countries are going to pay an enormous price," Rubin warned, He noted that rapidly advancing AI capabilities could bring serious consequences, including cyberattacks, adding that growing availability of open-source AI tools could put powerful capabilities in more hands, while continued advances in AI could increase uncertainty about future risks.

He suggested that a framework for cooperation should focus on containing AI-related risks while leaving room for both countries to benefit from the technology.

He did not outline specific provisions for a potential agreement, emphasizing the need for intensive engagement within each country and between them.

The remarks come as Beijing and Washington agreed to set up channels for dialogue on AI-related issues.

In September, both countries reached eight deliverables and understandings, including an agreement to establish a China-US AI Dialogue to exchange views on the risks and benefits of AI.

They agreed that the next exchange would take place in November 2026 and to establish a bilateral communication channel for AI-related incidents.