Why AI Governance is a Marathon: Insights from Prof. Ji Weidong, Co-Chair of the UNU Global AI Network Board

The UNU Global AI Network Insight Series

This blog marks the opening entry in the UNU Global AI Network Insight Series, launched to celebrate the Network’s two-year anniversary.

In this series, members of the UNU Global AI Network will share diverse insights at the intersection of artificial intelligence and sustainable development. Bringing together perspectives across regions and disciplines, the series aims to explore both the opportunities and challenges of shaping AI for all.

Launched at the UNU Macau AI Conference 2024, the UNU Global AI Network was established as a global platform to advance the vision of the Global Digital Compact through multi-stakeholder cooperation on responsible AI governance and digital development.

Learn more about the UNU Global AI Network at unu.edu/AINetwork.

What happens when artificial intelligence evolves faster than the institutions designed to govern it? Can legal systems built for the industrial era still contain technologies capable of autonomous reasoning, cross-border decision-making, and self-directed action? And as the race toward superintelligence accelerates, who is building the guardrails?

These questions framed Professor Ji Weidong’s keynote reflections at Tech Week Shanghai 2026. As Co-Chair of the UNU Global AI Network Board and Distinguished Chair Professor of Humanities and Social Sciences at Shanghai Jiao Tong University, Professor Ji argued that the world has entered a new phase of AI development: one in which technological acceleration is rapidly outpacing institutional adaptation.

His message was clear: AI governance cannot succeed through fragmented reactions, short political cycles, or isolated national regulations. Governing superintelligence will require a long-term global institutional transition grounded in coordination, adaptability, and sustained international cooperation.

Here is why AI governance is a marathon.

The Governance Gap Is Growing Faster Than the Technology

The urgency of this challenge was underscored by AI pioneer Geoffrey Hinton at the Digital World Conference at the UN Headquarters in Geneva. Hinton warned that only around 1% of global AI R&D funding is directed toward safety research, while the remaining 99% accelerates capability development - creating, in his words, “a very fast car with no steering wheel.”

The imbalance is amplified by the scale of investment driving AI deployment. According to Stanford University’s 2026 AI Index Report, global corporate investment in AI reached $581.7 billion in 2025. Technological capabilities are advancing at a geometric pace, while regulatory institutions, legal systems, and public understanding struggle to keep up.

What was once a theoretical concern has become a structural reality. The concept of “accelerationism” - the idea that technological growth outpaces legal and political systems - has moved beyond academic debate into state strategy and physical infrastructure. Mega-scale initiatives such as the $500 billion Stargate Project exemplify a new convergence of state power, computational capacity, and corporate influence.

Autonomous AI Is Already Stress-Testing Existing Institutions

The year 2025 was widely described as the “Year of the AI Agent.” Large language models evolved from passive conversational tools into autonomous systems capable of multi-step reasoning, cross-platform interaction, and independent task execution.

Autonomous AI agents can now navigate digital ecosystems with minimal human intervention, interact across institutional boundaries, and access sensitive information at unprecedented speed and scale. Real-world incidents have already demonstrated how automated systems can bypass security procedures, access financial information, or disrupt digital platforms.

The 2026 AI Index Report noted that global AI-related safety incidents increased by more than 55% year-on-year in 2025.

This transition fundamentally changes the risk landscape. The challenge is no longer simply whether AI systems are intelligent. The challenge is whether existing governance structures can respond quickly enough to systems that increasingly operate with autonomy, interoperability, and strategic agency.

Three Global Governance Models Are Emerging, But None Are Sufficient Alone

Different regions have responded with distinct governance architectures, each reflecting different political traditions and strategic priorities.

The European Union: Regulate First, Innovate Carefully

The European Union has built one of the world’s most comprehensive “hard-law” governance systems through frameworks such as the GDPR, Digital Services Act (DSA), and Digital Markets Act (DMA).

This model has established influential global standards for data protection and platform accountability. However, the same regulatory rigor that strengthens oversight can also increase compliance costs and reduce the agility of smaller digital enterprises.

The United States: Accelerate Innovation Through Flexible Governance

The United States continues to favor a more decentralized and innovation-oriented “soft-law” model that prioritizes market flexibility and voluntary corporate compliance.

This model has supported rapid technological development, but it has also produced regulatory fragmentation while giving major technology firms significant influence over policy direction. At the same time, frameworks like the Clarifying Lawful Overseas Use of Data Act (CLOUDA) and the Data Free Flow with Trust (DFFT) initiative reflect growing geopolitical efforts to secure strategic control over data flows.

China: Build Structured Flexibility Around Data Governance

China has pursued a governance model that combines foundational legal safeguards with flexible operational mechanisms intended to enable both security and data circulation.

This framework is anchored in three major legislative pillars: the Cybersecurity Law (2017), the Data Security Law(2021), and the Personal Information Protection Law (2021), alongside institutional innovations such as data exchanges, algorithmic filing systems, and the “Data 20 Rules.” The broader objective is to create baseline safety guarantees while enabling the controlled activation of data markets.

The Real Governance Challenge Is Structural, Not Merely Ethical

Professor Ji argued that future governance debates cannot rely solely on abstract discussions of “AI ethics” or universal value alignment.

Moral values, social norms, and political traditions differ significantly across civilizations. Attempting to impose a singular global ethical framework on advanced AI systems risks becoming both conceptually fragile and geopolitically contentious. Instead, governance efforts should increasingly focus on procedural and technical “goal alignment”: ensuring that AI systems remain predictable, auditable, controllable, and structurally safe across jurisdictions.

This reframing shifts the emphasis away from enforcing universal moral consensus and toward building interoperable governance mechanisms that can function across diverse political and cultural systems.

From Fragmented Regulation to a Global “Tripartite Governance Framework ”

If existing governance systems are increasingly fragmented, what could a more durable framework look like? Professor Ji proposed moving beyond reactive data localization measures toward a more coordinated international architecture built around a functional “tripartite governance framework ” for AI governance.

Governments Need to Establish the Legal and Institutional Foundations

Governments remain responsible for defining enforceable legal baselines.

This includes establishing safety thresholds for cross-border data transfers, maintaining critical data classifications and negative lists, enforcing algorithmic accountability systems, and creating robust regulatory mechanisms capable of real-time oversight.

But governance cannot remain purely reactive. Governments must also collaborate with enterprises to build testing environments, regulatory sandboxes, and auditing systems that evolve alongside technological capabilities.

Enterprises Need to Translate Governance into Engineering Practice

Enterprises are no longer simply market actors; they are operational architects of the AI ecosystem.

This places growing responsibility on companies to embed governance directly into technical design, internal auditing systems, compliance workflows, and safety engineering processes.

Professor Ji emphasized the importance of internationally recognized technical standards, including ISO frameworks and NIST guidelines, as mechanisms for building shared operational trust across jurisdictions.

International Organizations Need to Coordinate Global Interoperability

No single state can govern superintelligence alone.

International organizations therefore play a critical coordinating role: harmonizing standards, reducing enforcement fragmentation, facilitating cross-border certification systems, and advancing global digital literacy.

Universities and multilateral institutions also have an increasingly important role in cultivating the next generation of AI governance professionals capable of bridging technology, law, diplomacy, and public policy.

AI Governance Requires Historical Patience

One of Professor Ji’s most important reminders was historical.

Major technological revolutions have always required long periods of institutional adaptation. The legal and corporate systems that emerged after the Industrial Revolution took decades to stabilize. The governance of superintelligence will likely demand a similar process of experimentation, negotiation, and institutional reconstruction.

This is why framing AI governance as a race can be misleading.

Speed matters. But endurance, coordination, and institutional resilience matter more.

The future of AI governance will not be determined by whichever country or company moves first. It will depend on whether humanity can build governance systems capable of preserving human agency while sustaining an open, secure, and cooperative global digital order.

AI governance, Professor Ji argued, remains a marathon, not a sprint.

References & Sources

《2026年人工智能指数报告》（Artificial Intelligence Index Report 2026）， 报告网址：ai_index_report_2026.pdf。Sha Sajadieh, Loredana Fattorini, Raymond Perrault, Yolanda Gil, Vanessa Parli, Lapo Santarlasci, Juan Pava, Nestor Maslej, Russ Altman, Erik Brynjolfsson, Carla Brodley, Jack Clark, Virginia Dignum, Vipin Kumar, James Landay, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Elham Tabassi, Russell Wald, Toby Walsh, Dan Weld. “The AI Index 2026 Annual Report,” AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, Stanford, CA, April 2026.

《2026年人工智能指数报告》（Artificial Intelligence Index Report 2026），报告网址：ai_index_report_2026.pdf。（报告整理来源：启元洞见：https://mp.weixin.qq.com/s/RGk7FAZ0xAASXdletFZihw）

Cf. Nick Land, “Meltdown” delivered at the “Virtual Futures” conference in 1994,and included in his collection of papers Fanged Noumena: Collected Writings 1987-2007 (ed, by Raymond Brassier), London: Urbanomic, 2011.

《季卫东教授应邀在2026全球数字合作交流会暨全球数据周发表主旨演讲》，法的社会视野： https://mp.weixin.qq.com/s/3pb72SqjP0arxdoTsPFFkg

《联合国大学全球人工智能网络理事会联席主席季卫东：人工智能治理是场马拉松》，广州日报新花城： https://huacheng.gz-cmc.com/pages/2026/05/06/eeedfb31888e44b98001802c4fa832a0.html

The UNU Global AI Network Insight Series is launched to celebrate the Network’s two-year anniversary. In this series, members of the UNU Global AI Network share diverse insights at the intersection of artificial intelligence and sustainable development. Bringing together perspectives across regions and disciplines, the series explores both the opportunities and challenges of shaping AI for all.

The views expressed in published articles of this series are solely those of the author(s) and do not necessarily reflect the positions of the United Nations University, UNU Macau, the United Nations, or affiliated entities. Publication does not imply endorsement or institutional approval.

Authors are responsible for the accuracy and originality of their contributions. UNU reserves the right to edit, decline, or remove submissions in line with the purpose and values of the series.

For inquiries, please contact: AINetwork@unu.edu