Publications
Global Center for Development and Strategy
Publications
A Blueprint for Multinational Development
- 관리자
- 2025.12.08.
- 437
The research team led by Professor Kyung Ryul Park at the Center for Science, Technology and Global Development (G-CODEs) has co-published a policy report titled “A Blueprint for Multinational Advanced AI Development,” in collaboration with researchers from Mila – Quebec Artificial Intelligence Institute (Canada), the University of Oxford, RWTH Aachen University (Germany), the Technical University of Munich (TUM), and École Normale Supérieure (ENS-PSL, France).
This report analyzes the necessity and feasibility of establishing a multinational AI research and development cooperation body among *AI Bridge Powers, including South Korea, within the current US-China-centered global AI hegemony competition structure. It proposes strategic directions to simultaneously secure national sovereignty, economic competitiveness, technological innovation, democratic values, and national security through this cooperation.
*AI Bridge Powers: Second-tier technological powers such as South Korea, the UK, Canada, Singapore, Germany, and France
● Key Arguments
In the emerging bipolar frontier AI order, where advanced AI capabilities are concentrated in a small number of “frontier AI states,” AI bridge powers can feasibly develop frontier-competitive AI models through a multinational advanced AI partnership that pools compute, talent, data, and governance capacity. Such a partnership is essential to safeguard ▲National sovereignty, ▲Democratic values and legitimate governance, ▲Economic competitiveness and growth, ▲Technical innovation capacity, and ▲Long-term national security.
This approach enables bridge powers to avoid the structural vulnerabilities associated with exclusion from frontier AI development while preserving strategic autonomy.
● Global AI Competitive Landscape
- (Structural Concentration of Frontier AI Capabilities) AI compute, development talent, high-value data, and ownership of frontier models are highly concentrated, with the United States controlling roughly three-quarters of global AI compute capacity and China a large additional share. This concentration creates barriers that individual mid-sized economies cannot overcome through national efforts alone, making independent frontier AI development structurally unrealistic for most bridge powers.
- (AI as a Foundational Strategic Asset) Frontier AI systems are becoming deeply embedded in economic, scientific, cultural, and security infrastructures. Access to frontier AI increasingly shapes national productivity, innovation, and defense capability. States unable to develop or reliably access such systems risk enduring capability gaps relative to frontier AI states.
● Strategic Dilemmas Under a Bipolar AI Order
Bridge powers face a structural choice between dependency and weakness:
- (Dependency) Adopting U.S. or Chinese frontier AI systems creates structural dependence. Frontier AI states may exploit this position through privileged access to sensitive data, the ability to restrict or degrade services, selectively withholding advanced capabilities, embedding external values and design choices in foundation models, and shaping terms of trade to their advantage.
This undermines sovereignty and constrains autonomous policy and industrial strategies.
- (Weakness) Conversely, limiting the adoption of frontier systems to avoid dependency risks falling behind as frontier states use AI to achieve breakthroughs in economic productivity, scientific discovery, and military capability. These widening capability gaps can translate into long-term economic and strategic weakness.
● Policy Alternative: A Multinational Frontier AI Partnership of Bridge Powers
A multinational frontier AI partnership offers the only strategy that is both frontier-competitive and sovereignty-preserving at manageable cost.
- (Pooled Computing Infrastructure) By coordinating and pooling national and regional AI compute capacity, bridge powers can cover the high, fixed costs of frontier model training that no single member could sustainably bear alone. Shared infrastructure enables frontier-scale development without requiring each state to duplicate the full stack.
- (Pooled AI Talent Base) A large share of leading AI researchers have ties to bridge power countries. A well-resourced, values-driven multinational initiative can “call home” top talent by offering frontier-scale infrastructure, competitive conditions, and a mission grounded in responsible and legitimate AI development.
- (Pooled High-Value Data and Data Infrastructure) While much training data is public, high-quality expert-labeled and domain-specific data are scarce and costly. Joint data cleaning, labeling, and sharing arrangements among members create scale and differentiation advantages for frontier development.
- (Strategic Positioning in Trustworthy and Responsible AI) Bridge powers can leverage strong data protection regimes, rule of law, and accountable governance to lead in reliable, trustworthy AI, addressing unmet market demand in high-value and sensitive sectors. This positions the partnership as a trusted broker in global AI ecosystems.
● Expected Outcomes
Participation in such a partnership would:
- Preserve technological sovereignty and reduce exposure to coercive dependency
- Strengthen strategic autonomy in economic, technological, and security policy
- Enhance domestic industrial competitiveness through privileged access to frontier models
- Increase influence in global AI governance, shifting members from rule-takers to rule-makers
- Support democratic legitimacy and social stability through responsible AI stewardship
- Reinforce the long-term foundations of national security by narrowing capability gaps with frontier AI states
Even short of absolute frontier leadership, collective development significantly reduces the risks of dependency and weakness while expanding diplomatic, economic, and security leverage.
