CGAIG Visiting Scholar
Antonio Postigo (Spain)
Senior Fellow at the Barcelona Institute for International Studies (IBEI)
Preface
On the morning of May 21, Dr. Antonio Postigo, Visiting Scholar at the Center for Global AI Innovative Governance (CGAIG) and Fudan Development Institute (FDDI), Senior Fellow at the Barcelona Institute for International Studies (IBEI), and Visiting Research Fellow at the United Nations University Institute on Comparative Regional Integration Studies (UNU-CRIS), held an in-depth conversation with Dr. Xin Yanyan, Deputy Secretary-General of CGAIG and Assistant Professor at FDDI.
The interview took place during an important political moment following the recent meeting between the leaders of China and the United States. Before the summit, there had been considerable expectations that both sides might discuss some of the challenges associated with AI development and governance. Although no concrete agreements were announced, subsequent comments suggested that both countries recognized the importance of maintaining channels of dialogue on AI-related issues, including frontier-model risks, safety guardrails, and crisis communication mechanisms. At the same time, continuing geopolitical tensions between the two major AI innovation powers make it difficult to imagine the rapid emergence of a single, unified international treaty through traditional multilateral mechanisms such as the United Nations or other global institutions. In parallel, the digital divide and development gaps affecting the Global South, together with the emphasis placed by advanced economies on security and risk containment, reveal important structural divergences in global AI governance. Against this backdrop, Dr. Postigo drew on his background in international political economy and his previous experience working for the United Nations to analyze the significance of China-U.S. dialogue, the core concerns of the Global South, the potential advantages and limitations of Free Trade Agreements (FTAs) and Digital Economy Agreements (DEAs) as regulatory frameworks for selected dimensions of AI governance, and the demonstration effects of the Asia-Pacific region as a site of institutional experimentation.
The following is an excerpt and summary of Dr. Postigo’s core insights.
Dr.Xin Yanyan (Left) and Dr.Antonio (Right)
Photo by Li Ziyi（Research Assistant of CGAIG）
I. The Significance and Realistic Limitations of Restarting the China-U.S. AI Dialogue
Artificial intelligence is no longer only a commercial technology; it is increasingly linked to economic competitiveness, cybersecurity, and national security. When the two leading AI powers stop communicating, the risks of fragmentation in the international digital order and strategic mistrust increase significantly. For this reason, the restart of the China-U.S. AI dialogue carries both symbolic and practical significance. Even a limited dialogue is important because it creates basic channels for information exchange and risk management. To address extreme scenarios, such as advanced models becoming difficult to control or weaponized AI systems being misused by non-state actors, China and the United States could consider establishing a dedicated crisis communication channel, comparable in function to the “Red Phone” hotline established between the United States and the Soviet Union during the Cold War.
The fact that no immediate agreements were announced after the summit suggests that this meeting may have followed a different logic from summits where leaders formally sign agreements that have already been negotiated for months by lower-level officials. In this case, the process appears to have moved in the opposite direction. Given the strategic trust deficit between the two powers, the first step was not necessarily to reach a detailed agreement, but to signal political willingness to keep talking and to create a minimum basis for confidence-building. This may open space for more specific follow-up discussions at ministerial or technical levels, where concrete areas of cooperation could be explored in greater detail. In this sense, high-level diplomatic goodwill may help reopen institutional channels.
However, important limitations to this renewed China-U.S. AI dialogue remain. Statements after the summit by representatives of the current U.S. administration, including U.S. Treasury Secretary Scott Bessent, reflect a realist theoretical understanding of AI as an arena of strategic competition, technological leadership, and relative power, rather than primarily as a domain of cooperation. This perspective is also visible in the language of the current U.S. AI Action Plan, which frames national AI strategy around the idea of “winning the race.” From an International Relations perspective, this realist framing connects AI governance to economic competitiveness, national security, and geopolitical influence. However, it differs from institutionalist theories of international cooperation and from multilateral approaches, including those emphasized by China, which stress dialogue, coordination, Global South inclusion, and shared mechanisms for managing transnational risks. The challenge is therefore how to create channels of cooperation between the two leading AI powers despite strategic rivalry, different governance preferences, and competing visions of the international order.
Some scholars, including ÓhÉigeartaigh and Roytburg, have argued that, even under increasing geopolitical tensions, the framing of AI development as a zero-sum “race” between the U.S. and China should be treated with caution, particularly when this framing is used in U.S. policy debates. From this perspective, the “AI race” operates not only as a description of technological competition, but also as a constructed political narrative that may help justify reduced regulatory oversight and the prioritization of corporate and strategic interests over broader social and environmental concerns in the U.S. This concern has become more visible in recent legislative debates in the U.S. over the infrastructure required for advanced AI systems, including energy-intensive data centers and the fast-tracking of nuclear power or other energy projects to support AI development. In any case, whether the AI race is primarily an objective technological reality, a constructed political narrative, or some combination of both, zero-sum rhetoric among leading powers can undermine the prospects for international cooperation.
II. Core Concerns of the Global South and China’s Supply of International Public Goods
Future global AI governance should avoid being shaped primarily by the priorities of a limited group of advanced economies and should instead balance the safety and strategic-stability concerns that dominate many advanced-economy debates with the development and inclusion priorities of the Global South. The greatest concern for many developing countries is being excluded from the rule-making process and becoming passive rule-takers rather than active participants in shaping global AI governance. While much of the debate in advanced economies focuses on regulation, ethical accountability, misinformation, frontier-model risks, and long-term safety, many countries in the Global South view AI primarily as a tool for economic modernization, productivity growth, public-service delivery, and poverty reduction. They therefore prioritize concrete applications in health care, education, agriculture, financial inclusion, and industrial upgrading. If global governance frameworks focus too narrowly on risk containment through overly restrictive safety regulations, they may unintentionally widen technological gaps and deepen existing global inequalities. The 2025 Paris AI Action Summit illustrated these tensions: its final statement on inclusive and sustainable AI was supported by China and the EU, while the U.S. and the U.K. declined to sign it amid concerns over regulatory overreach, national interest, and the balance between innovation, security, and global governance.
China is strategically positioned to play a key role toward more inclusive AI governance debates, particularly by supporting practical forms of cooperation with developing countries. In response to a question on how China could contribute to the provision of international public goods for AI development and adoption in the Global South, Dr. Postigo noted several possible areas of contribution. These could include support for digital infrastructure investment, AI capacity-building, technical training, open-source cooperation, affordable cloud services, AI applications for development, and broader participation of developing countries in international governance discussions. Such contributions would be especially valuable if they help developing countries strengthen their own capacity to use, adapt, and govern AI in line with their domestic development priorities.
III. Trade-Based Frameworks in AI Governance and Demonstration Effects of FTAs/DEAs in the Asia-Pacific
While a broad multilateral framework—ideally with wide participation and treaty-based commitments—would in principle provide the most legitimate foundation for global AI governance, it is unlikely that a single comprehensive treaty can quickly address all dimensions of AI. Historically, especially after World War II, many global commons were governed through multilateral treaty-based regimes. However, since the end of the Cold War, global governance has also relied increasingly on more flexible, issue-specific, and network-based mechanisms. In the case of AI, rapid technological change, regulatory divergence, and geopolitical tensions among major powers create additional hurdles for a rapid multilateral agreement. AI governance is therefore likely to require a multi-layered approach. High-level bilateral dialogues, especially between China and the U.S., are important for strategic stability, military implications, existential frontier-model risks, and crisis communication. By contrast, FTAs and DEAs operate at a different institutional level, addressing more granular economic, regulatory, and developmental dimensions such as cross-border data flows, interoperability, technology diffusion, capacity-building, and digital infrastructure standards coordination.
Compared to non-binding or soft-binding multilateral declarations, FTAs and DEAs offer certain institutional advantages, including formalized commitments, bundled economic incentives, and a flexible “bottom-up” environment for experimentation. Because regulatory models vary significantly across borders, FTAs and DEAs allow smaller groups of countries to test rules and cooperation mechanisms incrementally. However, these agreements also have significant limitations. Although FTAs and DEAs may include dispute settlement procedures, monitoring mechanisms, and economic incentives linked to market access, these tools are primarily designed for trade and regulatory commitments. They are not well suited to deterring states or non-state actors from pursuing weaponized AI systems, uncontrolled frontier models, or other severe security risks. The main contribution of FTAs and DEAs is to support alignment in selected domains such as digital trade, data governance, standards cooperation, interoperability, capacity-building, and institutional dialogue. Once tested, some provisions may serve as modular building blocks for broader or complementary governance frameworks.
The Asia-Pacific is playing an important demonstration role in the development of trade-based digital and AI governance. As of May 2026, among the 20 trade agreements containing provisions that explicitly mention AI, 16 are located in the Asia-Pacific region. This regional concentration can be explained by at least three factors. First, Singapore has acted as an institutional catalyst in digital trade governance. As a highly open, services-oriented, and digitally connected economy, Singapore has been a central participant in many of the DEAs concluded to date, including DEPA, and has promoted innovative behind-the-border rules on cross-border data flows, paperless trade, digital identities, interoperability, and emerging technologies. Second, the Asia-Pacific is home to some of the world’s most integrated cross-border production networks and digital value chains, with China playing a central role in many of these structures. This creates strong incentives for governments to coordinate rules that facilitate data, services, investment, technology, and production linkages. Third, the Asia-Pacific is institutionally diverse, bringing together advanced, emerging, and developing economies with different regulatory capacities and development priorities. Such diversity could hinder cooperation and regulatory coordination, but the region has also developed pragmatic and flexible forms of cooperation, including ASEAN-centered regionalism, that support modular frameworks, minimum common denominators, and gradual rule development. This is relevant for AI governance because many developing economies may not be willing or able to adopt demanding AI regulatory models immediately, but they can still participate in cooperation on digital infrastructure, data governance, standards, capacity-building, and responsible AI adoption at a pace consistent with their domestic capacities.
Conclusion
The future architecture of global AI governance is likely to be a multi-layered ecosystem in which hard-binding frameworks, soft-law norms, technical standards, trade-based mechanisms, and diplomatic dialogues interact through gradual institutional learning. A broad multilateral framework with wide participation and treaty-based commitments remains an important long-term goal. However, expecting a single comprehensive multilateral organization to address all regulatory conflicts and severe AI risks is unrealistic in the current geopolitical context. Recent disagreements at international AI summits over inclusivity, sustainability, regulation, and security illustrate the difficulty of building consensus. Against this background, Dr. Postigo argued that China can play a leading role in promoting more inclusive AI governance through practical cooperation with developing countries, including digital infrastructure, technical assistance, capacity-building, and support for meaningful Global South participation in AI governance debates. He also emphasized that FTAs and DEAs cannot replace multilateral AI governance, but can provide practical building blocks in selected domains such as digital trade, standards cooperation, institutional dialogue, and technology diffusion. These incremental and regionally grounded mechanisms may help reduce technological exclusion and contribute to a more balanced and inclusive global digital order.
