US and China need shared risk vocabulary to manage AI threats: Beijing-linked think tank

Beijing and Washington should first develop a common language around AI risks rather than seek to resolve their broader technology rivalry, a leading Chinese think tank has advised.

The call from the China Institutes of Contemporary International Relations (CICIR), affiliated with the Ministry of State Security, comes days after Chinese President Xi Jinping and US President Donald Trump agreed to formalise a new AI dialogue mechanism during their summit in Washington, their second such meeting this year.

Under the agreement, both sides will exchange views on the risks and benefits of AI, as well as establish a bilateral communication channel for AI-related incidents, with the next dialogue due to be held in November.

Liu Chong, director of CICIR’s Centre for Sci-Tech and Strategic Studies, said the dialogue should begin by building a shared understanding of emerging AI risks and gradually establishing trust between the two sides.

01:52

Xi says China and US have responsibility to ‘manage AI for good’

In an article published on Monday, Liu wrote that the rapid development of artificial intelligence (AI) meant there was unlikely to be a single point at which technology could be clearly divided into safe and unsafe categories.

Instead, he proposed developing a dynamic system of “risk changes and response levels” compared with searching for an abstract “red line”.

Given the fluid nature of “superintelligence risks”, effective international governance must prioritise fostering mutual knowledge over pushing for comprehensive rule books, Liu argued, using the preferred Trump administration term for AI.

“This is where China and the US first need to get on the same page,” he wrote.

A White House statement following last week’s summit said the leaders had agreed to use the term “superintelligence” rather than “artificial intelligence” for the emerging technology under discussion. Beijing has said it respects the US’ choice of terminology and that further communication will be needed as the technology evolves.

Liu also suggested an approach to governance that respected sovereignty over technology while maintaining dialogue.

“China and the US do not need to adopt the same AI industrial policies, nor do they need to share model parameters and core technologies,” he wrote.

Lin suggested that they should rather develop a set of risk terms that could be understood by governments, companies and researchers on both sides. This would include common understandings of what constituted dangerous autonomy, such as when model behaviour should be treated as a security incident and when training should be slowed or halted.

CICIR’s call for a shared understanding comes as the two countries continue to compete over advanced technology, with Washington tightening restrictions on Beijing’s access to advanced chips and AI capabilities.

SCMP Plus is a new premium news platform that gives you an

all-inclusive edge to stay ahead on China news.

To access our exclusive content you’ll need to subscribe.

Already a subscriber?

LOG IN

It also called for stronger oversight of frontier AI companies, arguing that a small group of leading firms “should not determine” the direction of AI governance.

Liu said US frontier models were largely closed-source and that the lack of transparency had increased concerns outside the companies.

“US companies should take the lead in disclosing information about AI risks,” he wrote.

This month, Anthropic CEO Dario Amodei argued that the US should maintain its technological lead over China and continue restrictions on the sale of advanced AI chips and chipmaking equipment to Beijing.

Chinese officials have rejected such an approach and called for international cooperation on AI instead.

Liu acknowledged the tension, writing that the central challenge for AI governance was whether Beijing and Washington could preserve enough “trust space” to cooperate while their technological competition continued or intensified.

“The goal of dialogue should be to establish a minimum framework for managing risks that could survive even as the US-China technological competition continues or intensifies,” he wrote.

Beyond managing risks between Beijing and Washington, broader international access to AI safety tools and capabilities as part of global AI governance was needed, he said.

As AI advances, mature and widely applicable capabilities should be converted into accessible public goods, Liu added.

“The governance of superintelligence must look beyond merely curbing the spread of offensive risks,” he wrote. “It must also actively promote the adoption of defensive safeguards.”