China's AI Governance: Engaging the Global South

In June, Xi Jinping delivered a speech at the 60th anniversary of the UN Conference on Trade and Development calling for a truly multilateral approach to AI governance. He also reaffirmed China’s commitment to the global South. What is China’s strategy for promoting its approach to AI governance, and does it have a strategy specifically for engaging the global South on this issue?

Chinese strategy for AI governance on the global stage involves four key components. The first is that the PRC is pushing for a people-centered approach to developing AI, or what its leaders call yi ren wei ben, which has come up a lot in related policy documents that the PRC has published. The second emphasizes national sovereignty, which includes things like opposing the use of AI for ill purposes, manipulation of public opinion, and disinformation. The term national sovereignty is also linked to previous rhetoric on territorial integrity and digital sovereignty. The third component is related to a call for the establishment of testing and assessment of AI systems to prevent AI risks. The fourth one calls for more representation of developing countries in global AI governance. It emphasizes the role of the United Nations as a channel for communication and also welcomes “strengthening of North-South and South-South cooperation.”

While this all sounds great on paper, there are some levels of skepticism as to how these goals are brought into reality, given China’s domestic approach to AI governance.

China has assumed an influential role within the global South. How does its promotion of its AI governance approach within international institutions align with the development goals of the global South?

In terms of how the PRC is engaging with the global South community, the priority right now is messaging, particularly on how to further promote AI for development, as well as the role that China can play in bolstering other countries’ AI innovation.

First, the PRC focuses on AI governance for innovation, meaning running faster. If you look at the Shanghai Declaration on Global AI Governance, there is a great deal of detail related to three components. The first is technical training, which includes calls for the establishment of cooperation platforms to facilitate technology transfer and commercialization. The second component focuses on data flows and calls for safeguarding high-quality data development with high-quality and high-level data security. That means promoting the free flow of data between countries and increasing cooperation to build datasets that can be trained to improve the quality of AI models. The final component concerns talent, which calls for more cultivation of AI professionals, more education and training, and improvement of AI literacy and skills around the world. So China is positioning itself as the leader and champion of education and AI training.

In addition to AI governance for development and innovation, the PRC is also focusing on AI governance as it relates to risk management, specifically how AI technologies are developed in a way that manages risk. Risks can be related to cybersecurity, disinformation, abuse, or misuse, among other threats. But China does not have a good track record on this issue, especially when one considers 5G. Huawei’s 5G equipment has massive security vulnerabilities with backdoors built in. There is definitely a need to read policy documents very closely and look at how they actually are implemented.

What role do China-led multilateral platforms, such as the Digital Silk Road Initiative and Shanghai Cooperation Organization, play in achieving the country’s AI governance strategy and promoting such a framework throughout the region?

China has already laid the foundation to build its own version of multilateral platforms. As you mentioned, the Digital Silk Road is a key example. China-led international forums are meant to be leveraged to promote the PRC’s ambition of AI governance in countries where there is little presence of the United States or its allies. The Digital Silk Road and related multilateral platforms have long played a really important role in the PRC’s export of technology and talent, as well as training and research collaboration and infrastructure investment in data and cloud-computing centers. And we have seen evidence of this in Southeast Asian and Latin American countries.

The decision document from the 2024 Third Plenum laid out economic policy for China. The PRC has renewed its interest in expanding international science and technology exchanges and collaboration as stated in that decision document. It also notes the implementation of the Belt and Road Initiative’s Science Technology Innovation Action Plan, which seeks to strengthen multilateral cooperation for AI. So Chinese AI governance strategy on the global stage does not happen in a vacuum.

To the PRC, AI development is related to surveillance, data governance, economic development, and its “going out” strategy. We have seen examples of AI enabling surveillance technology exported to Southeast Asia. In the realm of data governance, there is a growing global gap as the United States is pulling away from digital policy negotiations, and China hopes to fill that gap. The PRC is expected to create more opportunities for its companies to go out into different countries by negotiating data agreements and other ways to allow Chinese companies to localize in these countries. It is pretty clear that the promotion of the Chinese version of AI governance on the global stage is already built on top of existing channels.

How does China balance its AI governance interests with those of other major powers while maintaining its leadership in the global South?

In order to align its approach to AI governance with the development goals of the global majority, China is appealing to the region with a promise of economic benefits. First, it is calling for more applications of AI in different fields that are beneficial to society. Looking at the Global AI Governance Initiative, those fields of AI applications include industrial innovation, environmental protection, and energy management, among other areas. The second component involves exporting safe city and smart city technologies that are leveraging AI for things like turbocharging traffic, logistics, law enforcement, and surveillance that may be attractive to certain types of governments around the world. Lastly, China calls for the respect of national sovereignty. As mentioned earlier, this principle is linked to the previous rhetoric around digital sovereignty and territorial integrity. There is an underlying objective here that is different from the U.S. version of digital sovereignty.

Once again, looking at the Global AI Governance Initiative, China’s version of AI governance opposes using AI technologies for the purpose of manipulating public opinion. One can defensively argue that it is code for controlling public discourse. If you look at China’s track record of AI governance domestically, the economic appeal is really how China is balancing its AI governance interests with those of other major powers in the region. And if we zoom out a little into the broader context of strategic competition, the global South countries have increasingly leaned toward the United States as a security partner and China as an economic partner. For instance, we know that 51 countries signed on to the U.S. U.S. Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy, but China is not one of them. On the security front, China is not offering any clear commitments to mitigating military AI risks. So it is difficult to say that military AI risks are at the top of the country’s priorities in promoting its version of AI governance on the global stage.

Does China’s approach to AI governance and digital governance more broadly acknowledge the ethical considerations surrounding responsible AI use? Has the Chinese Communist Party implemented any safeguards to address the tendency for AI to reproduce biases?

Before talking about what China is doing on the global stage, we do have to understand how it is promoting AI governance with respect to AI ethics at home. That can inform how we view what China is doing on the global stage. How countries define AI ethics can have a significant impact on how the AI will actually work domestically before it is exported to other countries.

A document published by China’s National New Generation AI Governance Specialist Committee in 2021 essentially incorporated ethics into the entire life cycle of AI. The title of that policy is “Ethical Norms for New Generation AI Intelligence.” Essentially these ethical norms and guidelines are meant to promote fairness, justice, harmony, and security while also avoiding problems like biases, discrimination, privacy, and information leaks. But in terms of implementation, so far we have seen more evidence of an emphasis on the security component. Of all the components that are mentioned, security is at the forefront. Most recently, the Cyberspace Administration of China has been testing how well large language models can censor political sensitivities and topics that are related to Xi Jinping. And these large language models are supposed to embody “core socialist values.” Such testing is the latest evidence of how ethical norms are actually implemented to be aligned with such values.

The vision for the development of AI, with AI ethics in mind, also aligns with the Third Plenum decision draft I mentioned earlier. It calls for an improvement of the development of generative AI in the context of promoting the integrated management of news propaganda and online public opinion. When we consider what China might be doing in exporting its own version of AI governance, it is important to look at what the country is doing domestically. The global South community needs to pay attention to the actual implementation of policies on top of what China is saying on paper, just to gauge what the PRC really means by ethical and responsible AI.

There is a quote from the Global AI Governance Initiative that reads, “We should establish and improve ethical principles, norms, and accountability mechanisms for AI, formulate AI ethical guidelines, and build science and technology ethical reviews and regulatory systems.” While those goals seem nice on paper, what do they actually mean? What does it mean when AI developers are being told to censor parts of their database to train certain large language models? And we are only talking about generative AI here. We are still not getting into the larger context of AI systems in China. Domestic policies will be an area of significance when it comes to how the global South community should interpret what the PRC means by AI governance.

Ngor Luong is a Senior Research Analyst in the Center for Security and Emerging Technology at Georgetown University.

This interview was conducted by Charis Liu while a TFAS intern with the Technology and Geoeconomic Affairs group at NBR.