Enhancing Global Governance of Artificial Intelligence

Guests

Xue Lan, Dean of the International Governance Research Institute of Tsinghua University
Tang Shiqi, Dean of the School of International Relations at Peking University
Song Guoyou, Professor at the Institute of International Studies at Fudan University

Introduction

Currently, the world is undergoing significant changes, with artificial intelligence (AI) emerging as a strategic technology that is profoundly altering human production and lifestyle. However, the risks and challenges posed by AI have also garnered widespread attention. Improving global governance of AI is a common challenge faced by the international community. General Secretary Xi Jinping has pointed out the need to prioritize human-centered and benevolent AI, strengthen governance rules within the UN framework, and promote a green transformation to help developing countries better integrate into the digital, intelligent, and green trends. The 14th Five-Year Plan outlines the establishment of a governance framework for AI with broad participation from countries, aiming to build an equitable, trustworthy, diverse, and win-win global AI ecosystem, and support developing countries in enhancing their AI capabilities. This discussion invites guests to explore how to improve global governance of AI.

Characteristics and Challenges of Global AI Governance

Host:

In 2025, General Secretary Xi proposed a global governance initiative aimed at constructing a more just and reasonable global governance system. Compared to more mature global governance topics like maintaining multilateral trade systems and addressing climate change, what are the characteristics of global AI governance?

Xue Lan:

General Secretary Xi emphasized that the global governance initiative aims to build a more just and reasonable governance system. The characteristics of global AI governance stem from the rapid iteration of AI technology and its far-reaching impacts, for which the international community is not fully prepared in thought and action. This situation leads to unique characteristics in AI governance compared to more established topics.

For instance, issues like climate change have reached consensus during times of better international cooperation. In contrast, the explosive development of AI coincides with rising geopolitical risks, placing AI governance in a complex situation of great power competition from its inception. Some countries are engaging in “decoupling” and building barriers in technology development, talent cultivation, and data interconnectivity, attempting to monopolize competitive advantages through technological blockades, which disrupts the originally tightly-knit global research and industrial collaboration network. In this context, the core subjects of AI governance, primarily composed of technologically advanced countries, are not collaborators but competitors, undermining the trust necessary for cooperative governance and leaving the international community without a foundation for collaboration in addressing the risks and challenges posed by AI technology. Unilateralism and protectionism have led to a fragmented governance system, weakening the universal binding force of rules and plunging AI governance into a dilemma of “divide and rule.”

Tang Shiqi:

The pace of AI development far exceeds that of other fields, and its future holds immense uncertainty. Additionally, AI governance has two main characteristics:

AI is not only an object of decision-making but also partially participates in it. Due to its exceptional information processing capabilities, AI has become a tool for decision support in many areas. The reliance on AI by decision-makers will likely continue to increase, and the authenticity and effectiveness of the information provided by AI will have a growing impact on decision outcomes and quality. However, the AI technologies used for decision support are typically not controlled by decision-makers but by companies with significant advantages and powerful resource integration capabilities, which dominate foundational models and possess vast user data, thus wielding substantial influence. Therefore, ensuring that AI provides comprehensive and objective information has become an important issue.
The governance objects are fluid and virtual. Unlike trade interactions or climate change, which can be observed and controlled through tangible entities, AI governance involves computing power, algorithms, data, and models, which are technologies themselves rather than results produced by technologies. This lack of clear “governance anchors” complicates AI governance. Furthermore, AI exists and flows freely in cyberspace, largely beyond physical grasp and control. The information and data collected by AI, as well as the specific processing methods, are not entirely within human control. Generative AI possesses a closed-loop mechanism of “learning-optimizing-regenerating,” and its behavior may exceed the initial design of developers, leading to a disconnect between intention and outcome.

Song Guoyou:

Compared to relatively mature global governance issues, AI governance has at least three characteristics:

Uneven governance impact. Climate change is global, affecting all countries significantly. In contrast, the development and impact of AI are not uniform. Some countries have already felt the severe impact of AI, while others may perceive little or no impact.
Unpredictability of governance paths. As AI is still in its early development stage and has yet to be defined, it is challenging to implement specific governance focuses and rules, leading to unclear governance paths.
High sensitivity of technological competition. AI, as a key driver of a new round of technological revolution and industrial transformation, is a strategic high ground for great power competition. This high sensitivity to technological competition leads to strong competition and confrontational thinking in some countries, with insufficient awareness of cooperation and mutual benefit.

Challenges in Establishing a Global AI Governance Framework

Host:

China advocates for building a community with a shared future for mankind and a community of shared destiny in cyberspace, emphasizing the dual principles of development and security. The country has proposed the Global AI Governance Initiative, welcoming all stakeholders, including governments, international organizations, enterprises, research institutions, civil society organizations, and individuals, to collaboratively promote AI governance. What challenges does the current establishment of a global AI governance system face?

Xue Lan:

First, there is a lack of consensus in the international community on prominent issues in global AI governance. For example, how to recognize the potential risks of AI? What red lines need to be drawn in AI development? How can governments achieve a dynamic balance between promoting innovation and managing risks? Significant differences remain, posing challenges to the implementation of the Global AI Governance Initiative. Additionally, bridging the widening intelligence gap is an important issue that China and other developing countries hope to highlight for international attention.

Second, AI technology is in an explosive development phase, with technological breakthroughs often outpacing the establishment of governance rules. This “time lag” leads to governance always “chasing” technology, creating governance challenges. For instance, existing governance rules are often lacking or lagging in response to the risks posed by new technologies such as generative AI and embodied intelligence. Consequently, the agile governance methods advocated by China for AI safety governance have garnered significant attention from the international community.

Third, while there appear to be numerous governance mechanisms at the global level—such as the UN and the OECD promoting relevant guidelines—these mechanisms lack coordination, resulting in overlaps and even conflicts, creating a “mechanism complex” that complicates rules and makes implementation costly, significantly reducing actual effectiveness.

Tang Shiqi:

Currently, there is a rising tide of technological nationalism worldwide, with some countries prioritizing their technological security and industrial chain resilience over global public interests. From the corporate perspective, AI development is generally in a state of disordered competition lacking rules, intertwined with competition among countries, making it challenging to establish a global AI governance system. There are also numerous differences among countries regarding key issues such as cross-border data, content review, and government regulatory authority. Reducing conflicts and contradictions requires various international actors, especially great powers, to enhance mutual trust and cooperation, establishing a relatively complete AI monitoring and early warning mechanism, as well as a mechanism for sharing interests and responsibilities, to strengthen consensus as much as possible.

Song Guoyou:

From the perspective of the stakeholders involved in building a governance system, there are three difficulties:

Some countries pursue unilateralism and protectionism, hindering the construction of a global AI governance system.
Some countries with underdeveloped AI technologies lack enthusiasm for participation. These countries have not yet truly experienced the impacts of AI and lack awareness of its opportunities and risks, resulting in a lack of urgency to engage in global AI governance.
The private sector involved in building the AI governance system has concerns. Large AI companies, while being important subjects of global AI governance, are also objects of governance. These private sectors wish to center their development around profit to gain market-leading advantages and excess returns, leading to doubts and resistance towards a government-led AI governance system.

Human-Centered and Beneficial AI: Gathering Governance Forces

Host:

Under the current international mechanisms, AI governance often lags behind the iterative evolution of AI technology, leading to a widening global governance deficit. What governance concepts should be promoted globally to achieve a coordinated advancement of technological development and governance effectiveness?

Xue Lan:

First, we must adhere to a human-centered development philosophy. Ultimately, the development of AI should serve human society. We have consistently emphasized “technology for good,” making human well-being the fundamental goal of technological development. Whether in technology design or rule formulation, it should revolve around human needs, ensuring that technological development does not deviate from the trajectory of human social development, helping to bridge the development gap between the global north and south, and promoting shared benefits from AI development among countries.

Second, establish a governance foundation based on equal dialogue. Global AI governance cannot be dictated by a few technologically advanced countries. While countries differ in capabilities and sizes, they should participate equally in the formulation of governance rules, adhering to a global governance perspective of consultation, co-construction, and sharing. We need to reject unilateralism, build multilateral platforms, and fully respect the voices of different countries, especially the demands of developing countries, enhancing their representation and voice. Only through inclusivity and mutual learning can we truly form a governance framework with consensus.

Third, we must focus on action-oriented approaches, actively promoting the establishment of governance paths for inclusive development. Currently, there are significant technological gaps among countries. If this gap is allowed to widen, the governance divide will only deepen. Therefore, China emphasizes the need to strengthen capacity building, systematically plan, and promote overall progress through technical assistance and knowledge sharing, enabling global south countries to share in the dividends of AI development.

Finally, we need to build a risk prevention and control system for safe co-governance. AI risks are global; a technological flaw in one country could trigger a global chain reaction. Therefore, AI safety governance cannot be limited to “sweeping the snow in front of one’s own door.” AI safety should be treated as a global public good, with countries jointly developing safety technologies, establishing unified standards, and creating emergency cooperation mechanisms.

Tang Shiqi:

On this issue, I believe we must adhere to three points:

Human-centeredness. Traditionally, we believe that what makes us human is largely our rationality. The emergence of AI, which far surpasses human wisdom in certain aspects, particularly in logical reasoning, undoubtedly challenges humanity’s traditional status. Therefore, how we maintain human dignity, stimulate human creativity, and uphold basic humanistic bottom lines in the face of AI is a fundamental issue that requires collective exploration and consensus among all humanity.
Mutual benefit and cooperation. Currently, the development and utilization of AI reflect a multi-centered and decentralized nature, showing significant imbalances globally. Thus, on one hand, we need to complement each other’s advantages to avoid redundant construction and waste; on the other hand, we should ensure that countries and regions lagging in AI research and development can also enjoy the immense conveniences brought by AI, thus narrowing the digital divide. Both aspects need to be advocated in the process of global AI governance.
Openness and mutual trust. AI technology is based on large language models; the more open the information, the faster AI can develop. The application of AI in military and national security fields, along with its vast economic potential, makes it a key area in great power competition. Given that this situation is unlikely to fundamentally change in the short term, we should strive to find a balance in global AI governance, balancing national security, economic competition, and openness with mutual trust.

Song Guoyou:

The rapid development of AI technology necessitates that global AI governance ensures that AI technology, regardless of how it iterates, aligns with the common interests of all humanity, serving the well-being of all. Specifically, we should strengthen governance concepts such as open intelligence, inclusive intelligence, beneficial intelligence, and safe intelligence. Open intelligence means preventing certain major countries from creating AI camps and blocking others. Inclusive intelligence means that global AI governance should encompass the civilizations of all humanity, avoiding the injection of civilizational exclusivity into AI. Beneficial intelligence means ensuring that AI benefits all humanity, allowing countries worldwide to gain from AI technological innovation. Safe intelligence means actively preventing various risks and challenges that AI may pose to national security, ensuring it does not threaten humanity. Only by incorporating these four concepts into the entire process of global AI governance can we establish solid governance “guardrails” to achieve coordinated advancement of technological development and governance effectiveness.

Establishing a Broadly Participatory AI Governance Framework

Host:

China is hosting the 2025 World Artificial Intelligence Conference and the High-Level Meeting on Global AI Governance, releasing the “AI Global Governance Action Plan” and proposing to promote the construction of an inclusive and equitable multilateral global digital governance system. How should current international cooperation on AI break through geopolitical barriers and translate these consensus into actionable trust mechanisms and practical outcomes?

Xue Lan:

First, we must firmly support the UN in playing its primary role, advocating for the establishment of an “Independent International Scientific Group on AI” and a “Global Dialogue Mechanism on AI Governance” within the UN framework. At the same time, we should actively encourage and participate in various bilateral and multilateral dialogue mechanisms, forging a path of open consultation and diverse participation for mutual trust and governance. Second, while encouraging technological innovation and industrial application, we should establish an AI risk testing and assessment system through international cooperation, improve data protection norms, and seek a balance between development and security. Additionally, we should actively promote open cooperation, assisting countries worldwide, especially those in the global south, in enhancing their capacity and promoting the shared benefits of AI development.

Song Guoyou:

General Secretary Xi emphasized the need to “further expand AI cooperation, strengthen information exchange and technological collaboration, jointly manage risks, and form a governance framework and standards with broad consensus to continuously enhance the safety, reliability, controllability, and fairness of AI technology.” To break through geopolitical barriers in international AI cooperation, we can focus on the following:

Cooperation on important issues. We can first collaborate on significant issues that pose the greatest risks to humanity and have public product attributes, such as managing military applications of AI and developing AI for healthcare and disaster mitigation.
Engaging in pragmatic cooperation under existing multilateral mechanisms to enhance cooperation confidence, such as supporting the “Global Dialogue Mechanism on AI Governance” initiated by the UN and strengthening AI governance within the UN framework to promote relevant international cooperation.
Promoting bilateral and small multilateral cooperation as a beneficial supplement to international multilateral cooperation. Bilateral and small multilateral cooperation can be more targeted and flexible, addressing AI-related issues of common concern among members and achieving positive outcomes in AI governance while gradually expanding the consensus for cooperation.
Encouraging market cooperation among the private sector. Market cooperation is an important component of international cooperation; we should actively encourage the private sector to engage in various forms of commercial cooperation under the premise of complying with relevant regulatory rules, solidifying the cooperation foundation at the market level and creating cooperation models to break down geopolitical barriers and promote governance cooperation among nations from the ground up.

Host:

Currently, global south countries face dual challenges of an AI technology gap and insufficient governance discourse power. How can we ensure that global south countries participate equally in global AI governance?

Xue Lan:

The dual challenges faced by global south countries stem from weak educational and research foundations, a lack of talent support, insufficient technological capabilities, and outdated information infrastructure. In such circumstances, discussing technology empowerment becomes impractical. Therefore, to prevent the AI technology gap from widening further, we must pay close attention to the education and technological capacity building of global south countries, as well as their information infrastructure development, emphasizing “teaching them to fish.”

In terms of participation in AI governance decision-making, we need to strengthen the governance capacity of global south countries, which complements the practical application of AI technology. Inclusive AI development can enable global south countries to enjoy the benefits of AI while enhancing their capacity to participate in global AI governance.

Tang Shiqi:

General Secretary Xi pointed out the need to “strengthen international governance and cooperation on AI to ensure that AI serves humanity and avoids becoming a game for the rich and powerful.” We can promote and ensure the equal participation of global south countries in the AI governance process and allow them to enjoy the benefits of AI development from three levels:

Technical Level: In development issues, great powers have more room for cooperation, such as jointly constructing global public product-type AI systems, including carbon emission monitoring platforms for climate change, global infectious disease early warning models, and crop health management systems, and opening them to global south countries. At the same time, we should help global south countries establish their own AI systems, cultivate relevant talents, and build their development capabilities.
Rules Level: In the process of formulating rules for global AI governance, we should fully incorporate the participation of global south countries, ensuring fair distribution of representation to ensure that their demands can influence the basic direction of future AI development.
Value Level: On one hand, AI powers should avoid exporting AI systems that carry specific value systems to global south countries through cloud services or pre-trained models, thus shaping new forms of dependency; on the other hand, the future development of AI should fully reflect the cultural and value factors of global south countries. Since current AI development is based on human language systems, it is not entirely neutral in terms of values. The characteristics of civilizations are reflected in both value and ethical dimensions and in ways of thinking, all of which are embedded in a nation’s language system. An important treasure trove for future AI development lies in the language and cultural resources of non-Western countries. If these resources can be incorporated into AI’s corpus, it will significantly enhance AI’s wisdom and provide new possibilities for the development of global south countries and human civilization.

Song Guoyou:

First, global south countries should rely on themselves, promoting joint self-strengthening from both capability building and mechanism construction to break the structural asymmetry between the global south and north in AI governance, rather than waiting for or relying on developed countries to treat them equally.

Second, we should strengthen the AI capability building of global south countries to narrow the AI gap with developed countries. This can start with improving AI infrastructure, such as ensuring stable power supply and network connectivity, balancing the layout of computing power clusters across different regions in the global south, expanding the use of quality AI products, and reducing the costs of AI usage. We should also promote south-south educational cooperation in AI, establishing AI training centers to help global south countries enhance their AI research capabilities and cultivate top AI talents.

Finally, we should improve the mechanisms for global AI governance, enhancing the discourse power of global south countries in governance mechanisms. We can leverage important multilateral mechanisms such as the UN, BRICS, and the Shanghai Cooperation Organization to coordinate the common interests and governance demands of global south countries in the AI field, promoting the formation of a southern agenda in international AI governance. Global south countries should actively contribute important ideas to global AI governance, injecting their consensus-based ideas into the global AI governance process.