Can democracies cooperate with China on AI research?

Rebalancing AI research networks


China looms large in the global landscape of artificial intelligence (AI) research, development, and policymaking. Its talent, growing technological skill and innovation, and national investment in science and technology have made it a leader in AI.

Over more than two decades, China has become deeply enmeshed in the international network of AI research and development (R&D): co-authoring papers with peers abroad, hosting American corporate AI labs, and helping expand the frontiers of global AI research. During most of that period, these links and their implications went largely unexamined in the policy world. Instead, the nature of these connections was dictated by the researchers, universities, and corporations who were forging them.

But in the past five years, these ties between China and global networks for R&D have come under increasing scrutiny by governments as well as universities, companies, and civil society. Four factors worked together to drive this reassessment: (1) the growing capabilities of AI itself and its impacts on both economic competitiveness and national security; (2) China’s unethical use of AI, including its deployment of AI tools for mass surveillance of its citizens, most notably the Uyghur ethnic group in Xinjiang but increasingly more widespread; (3) the rise in Chinese capabilities and ambitions in AI, making it a genuine competitor with the U.S. in the field; and (4) the policies by which the Chinese state bolstered those capabilities, including state directed investments and illicit knowledge transfers from abroad.

Taken together, these concerns led to intense scrutiny and new questions about these long-standing ties. Is cooperation helping China overtake democratic nations in AI? To what extent are technologists and companies in democratic nations contributing to China’s deployment of repressive AI tools?

This working paper considers whether and to what extent international collaboration with China on AI can endure. China has been a subject of discussions among the government officials and experts participating in the Forum for Cooperation on AI (FCAI) over the past two years. The 2021 FCAI progress report identified the implications of China’s development and use of AI for international cooperation.The report touched on China in connection with several of the recommendations regarding regulatory alignment, standards development, trade agreements, and R&D projects but also focused on Chinese policies and applications of AI that present a range of challenges in the context of that nation’s broader geopolitical, economic, and authoritarian policies. A roundtable discussion on December 8, 2021 presented these issues to FCAI participants more fully and elicited their views.

This paper expands and distills this work with a focus on the scope, benefits, and prospective limits of China’s involvement in international AI R&D networks. In Part I, it presents the history of China’s AI development and extraordinarily successful engagement with international R&D and explains how this history has helped China become a global leader in the field. Part II shows how China has become embedded in international AI R&D networks, with China and the United States becoming each other’s largest collaborator and China also a major collaborator with each of the other six countries participating in FCAI. This collaboration takes place through multiple pathways: enrollment at universities, conferences, joint publications, and work in research labs that all operate in various ways to develop, disseminate, and deploy AI.

Part III then provides an overview of the economic, ethical, and strategic issues that call into question whether such levels of collaboration on AI can continue, as well as the challenges and disadvantages of disconnecting the channels of collaboration. The analysis then looks at how engagement with China on AI R&D might evolve. It does so primarily through a U.S.-focused lens because the U.S., as by far China’s largest competitor and collaborator in AI, provides an umbrella and a template for countries and FCAI participants that also collaborate with China on AI R&D and face many of the same issues. Moreover, measures to respond to the challenges China presents are more likely to be effective in coordination than in isolation. Recent U.S. export controls on semiconductors and the technologies used to manufacture them have laid bare the critical role of countries such as Japan and Korea. For now, the U.S. government is able to force foreign compliance through administrative measures, such as the foreign direct product rule, but these mechanisms may be made moot if foreign manufacturers engineer U.S. technology out of their supply chain. This paper deals with cooperative research rather than hardware supply chains, but similar dynamics exist across these domains. Accordingly, this paper is not just about collaboration with China but also about collaboration in relation to China.

Measures to respond to the challenges China presents are more likely to be effective in coordination than in isolation.

The U.S., other governments participating in FCAI, and their partners are not the only actors in this drama. What AI R&D with China looks like going forward will also be determined by what China does. China’s intensifying push for technological self-reliance has accelerated China’s disengagement from the international technology ecosystem in certain respects, while so far keeping it deeply enmeshed in other international research networks. The future trajectory of this engagement will depend heavily on actions taken by the Chinese government and the Chinese Communist Party.

In light of the issues presented by these changes, the paper proposes rebalancing AI R&D with Chinese researchers and institutions through a risk-based approach. Going forward, such collaboration will require a clear assessment of the costs and benefits, aiming to maximize the benefits of an open research environment and strong international links with the risks presented by AI R&D with China. Adopting an appropriately risk-based approach often will not counsel complete disengagement with China on AI R&D and instead require a rebalancing that takes into account the various vectors for knowledge transfer. Crucially, governments need to work collaboratively with each other and with companies, universities, and research labs to inform the assessment of the risks and understand the benefits of AI R&D with China. A failure to build these partnerships into the risk-assessment process could lead to bad outcomes that mismeasure risks and benefits, leaving the U.S. worse off.

Download the full report.


  • Footnotes
    1. Cameron F Kerry et al., “Strengthening International Cooperation on AI,” n.d., 123.