AI-DPI Geopolitics: Strategy for the Global Majority

Contents

Executive Summary

This panel discussion examines the collision between two distinct technology ecosystems: Digital Public Infrastructure (DPI)—characterized by openness, public ownership, and citizen benefit—and AI innovation—dominated by proprietary systems and competition between US and Chinese tech giants. The speakers argue that integration of AI into DPI is inevitable but must prioritize user needs, sovereignty, and public interest over competitive advantage, while acknowledging the real geopolitical constraints and resource limitations facing the Global South.

Key Takeaways

Integration is Inevitable; Governance is Critical — AI and DPI integration is already happening. The real question is not whether to integrate but under whose terms, with what safeguards, and at what cost to sovereignty and public interest. Proactive governance design is urgent.
User-Centric, Not Technology-Centric — Successful AI-DPI implementation requires flipping the priority: start with citizen needs and concrete use cases, then determine which technology (proprietary, open-source, hybrid) serves those needs—not the reverse.
Sovereignty Requires Capacity and Budget — Digital sovereignty is not theoretical. It requires: (a) domestic developer talent with deep platform knowledge, (b) government capacity to manage systems independently, (c) dedicated national budgets (not donor dependence), and (d) freedom from single-vendor lock-in.
Minimum Standards, Flexible Implementation — Global AI governance should establish a minimal, irreducible set of standards (anchored in human rights) and allow significant contextual variation. Overly prescriptive frameworks risk excluding countries from their own AI journeys or fostering non-compliance.
South-to-South Cooperation is Viable and Necessary — Rather than waiting for UN consensus or fearing US-China bifurcation, developing countries should actively form bilateral and regional cooperation frameworks, share implementation knowledge (e.g., via "third spaces" and research partnerships), and build indigenous expertise.

Conference Talk Summary

Key Topics Covered

DPI vs. AI: Competing Paradigms — Fundamental differences in operating principles, ownership models, and design philosophies
Integration & Governance — How AI and DPI can be integrated responsibly with appropriate safeguards and sovereignty protections
Global South Agency — Building autonomous pathways and capacity in developing countries rather than dependency on foreign tech stacks
Sovereignty & Interoperability Tension — Whether digital sovereignty is compatible with the interoperability principles central to DPI
International Cooperation Models — "Third spaces" and bilateral/regional collaboration as alternatives to geopolitical fragmentation
Funding & Financial Models — Role of philanthropic funding, multilateral development banks, and national budgets in sustaining DPI-AI infrastructure
UN Governance Mechanisms — Scientific panels, global dialogue, minimum standards, and capacity development networks
Contingency Planning — Preparing for geopolitical disruption and potential technology stack fragmentation
India's Leadership Role — India positioned as a "southwestern partner" offering an alternative development model
Use Cases & User-Centric Design — Emphasizing practical applications and citizen benefit over technological competition

Key Points & Insights

Two Technology Worlds Colliding: DPI operates on principles of transparency, public ownership, openness, and citizen benefit; AI operates on proprietary, competitive, commercially-driven models. These are now integrating into government services (chatbots, fraud detection, translation), raising critical governance questions.
User Needs as the Convergence Point: Both DPI and AI ultimately should be driven by user needs and concrete use cases rather than technological competition. The mismatch occurs when LLM development and foundational model competition (driven by US-China rivalry) takes precedence over practical citizen benefit.
Sovereignty is Ill-Defined: Across multiple panelists, "sovereignty" emerges as a fundamentally contested concept. It differs by country, by layer of the tech stack (data vs. compute vs. governance), and by context. Moving toward "digital sovereignty" or "data sovereignty" with nuanced definitions is more pragmatic than claiming absolute autonomy.
Interoperability ≠ Loss of Sovereignty: Contrary to common assumption, interoperability (DPI's core principle) can actually preserve sovereignty by preventing vendor lock-in, provided countries maintain decision-making autonomy and avoid dependency on proprietary systems.
Funding is Fundamental to Sovereignty: Ethiopia's decision to decline World Bank grants and develop its own digital ID system—and build internal government capacity rather than contracting to private system integrators—demonstrates that true sovereignty requires dedicated national budgets, not donor dependence.
Minimum Standards Over Complex Global Frameworks: Rather than top-down global AI governance regimes, the UN and member states should focus on identifying the irreducible minimum set of governance mechanisms (grounded in human rights) and allow countries to contextualize implementation.
The Multilateral Development Bank Problem: MDBs fund DPI/AI projects on project-based paradigms (capital expenditure) but inadequately support operational expenditure and long-term capacity building. Their approval timelines (2+ years) cannot keep pace with AI evolution.
India's Lived Experience Advantage: India has navigated the full lifecycle of DPI-AI integration (Aadhaar, UPI) with multiple vendor participation, public-private partnerships, and open-source-first approaches—lessons not yet well-documented or evangelized globally.
Africa's Immediate Priorities Trump Fragmentation Risk: While geopolitical fragmentation is a theoretical risk, most African countries face prior challenges: basic connectivity (only ~40% have internet), employment for youth, and immediate socioeconomic development. Technology discussions must be layered to address foundational infrastructure first.
South-to-South Cooperation as an Alternative: Rather than a "third way" (building entirely new tech stacks), cooperation among developing nations—through bilateral agreements, regional bodies, and collaborative standard-setting—may be more feasible and effective than attempting to exit existing tech ecosystems.

Notable Quotes or Statements

"DPI is not being built for the sake of DPI...DPI should bring some concrete use cases. AI on the opposite side is much more operating in a different life cycle starting from data...But the real implementation of AI when it comes to the citizens and population is driven by different use cases. That's where we have convergence." — Dr. Medi (UN Senior Adviser)

"The biggest issue is that the industrial and economic targets and goals are not well aligned with user needs...we are talking more about LLMs and foundational models than the real implementation of AI on the ground." — Dr. Medi

"Ethiopia declined a grant from the World Bank to develop their own vision of the digital ID. That's a notion of strong sovereignty signal." — Dr. Medi (on funding and autonomy)

"Sovereignty was a big discussion...if you dissect what sovereignty means for each country it differs. If it just means autonomy in decision-making and avoiding vendor locking, interoperability is actually the right principle to have, which preserves sovereignty rather than threatening it." — Surv Daz (Code Develop)

"A lot of the DPI momentum is also driven by donor funding. But more and more countries are starting to figure out how to fund DPI themselves...countries themselves are starting to think about how to fund DPI across sectors and functions as an infrastructure layer." — Surv Daz

"What Ethiopia did from the very beginning was they did not rely on any private sector system integrator...their government had some capacity of developers who took MOSIP, broke it down, started building it from scratch for themselves." — Surv Daz (on avoiding vendor lock-in)

"Rather than going for global governance huge frameworks, it's about thinking of what is the minimum standards, what is the irreducible set of policy rules and governance mechanisms that should be put in place." — Dr. Medi (on UN governance approach)

"Even before we start looking at disruption, the reality is how do you connect Africans? About 40% are using the internet. So you're adding another layer of fragmentation for policymakers...priorities have maybe not yet matured to where we're thinking about fragmentation." — Jane Monga (Carnegie Africa Program)

"India can do a better job of having that [institutional knowledge] out there...ultimately the idea is what problem are we solving for and what value are we creating for the citizen." — Surv Daz

Speakers & Organizations Mentioned

Identifiable Panelists

Dr. Medi — Senior Adviser to the UN Security General's Special Envoy for Digital and Emerging Technologies; member of UN Office for Digital and Emerging Technologies
Jane Monga — Africa Program (Carnegie; author of research on third-way AI development and regional cooperation)
Kunal — Principal at APT Institute
Surv Daz — Chief of Staff of Investments at Code Develop (philanthropic organization focusing on DPI partnerships)
Rowan (moderator)

Institutions & Organizations

United Nations (UN Office for Digital and Emerging Technologies; Security Council)
Code Develop (philanthropic organization)
APT Institute
Carnegie (Carnegie Institution; Carnegie Africa Program)
World Bank
Multilateral Development Banks (MDBs)
Indian Government (referenced for DPI leadership)
Ethiopian Government (Ministry overseeing FIA ID)
African Union and regional bodies (referenced)

International Initiatives/Agreements

Global Digital Compact (GDC) — UN framework announced in 2024 for digital cooperation
Scientific Panel on AI — UN-convened panel of 40 scientists delivering reports on AI risks/opportunities (modeled on IPCC climate panel)
Global Dialogue on AI — Multistakeholder mechanism under UN auspices; first session July 2025 in Geneva
Global Network on Capacity Development on AI — Member-state-led initiative

Countries & Regions Referenced

India (flagship DPI systems, third-way positioning)
Ethiopia (digital ID innovation, Fayda ID system)
Kenya (bilateral cooperation with India)
China & US (tech competition, foundational model race)
Brazil (latam GPT mentioned)
Estonia (comparison for small-nation DPI)
Africa / African continent (systemic focus)
Global South / Global Majority (primary audience)

Technical Concepts & Resources

DPI Systems & Platforms

Aadhaar — India's digital identity system (15+ years development; multi-vendor biometric architecture)
UPI (Unified Payments Interface) — India's interoperable digital payment system (open API, public-private partnerships)
MOSIP (Modular Open Source Identity Platform) — Open-source digital ID platform (adopted by Ethiopia for Fayda ID; used as example of government-managed implementation avoiding vendor lock-in)
Fayda ID — Ethiopia's digital identity system (built on MOSIP; entirely managed by Ethiopian Ministry; layering AI use cases)

AI Systems & Models

Large Language Models (LLMs) — Noted as focus of competitive development (US-China race) but criticized as misaligned with citizen use cases
Foundational Models — General-purpose AI models (referenced as commercially driven vs. use-case driven)

Governance & Standards

Digital Public Goods (DPGs) — Open-source software meeting certain technical standards; referenced as partial solution for AI governance but insufficient without procurement/governance frameworks
IPCC Model — UN Scientific Panel on Climate Change; referenced as model for proposed UN Scientific Panel on AI
Regulatory Sandboxes — Discussed as potentially valuable for collaborative standard-setting and testing across countries/sectors (though implementation challenges noted)

Data & Datasets

Masak / African voice — Coalition building large datasets for African populations (mentioned as example of regional data infrastructure)

Concepts

"Third Way" / "Third Pathway" — Development model distinct from US and Chinese tech stacks; discussed as aspirational but potentially impractical; reframed as south-to-south cooperation
"Third Spaces" — Proposed linkages between innovators in different developing regions (India, Brazil, etc.) to facilitate knowledge sharing and co-development
Digital Sovereignty / Data Sovereignty — Core contested concepts; acknowledged as context-dependent and layered (varies by country, by governance level, by data vs. compute vs. decision-making)
Vendor Lock-In — Risk of dependency on single technology provider; mitigation emphasized through multi-vendor participation and open-source adoption
Public-Private Partnerships (PPPs) — Integration model referenced as increasing in Global South; requires careful governance to maintain public benefit

Policy Frameworks

Minimum Standards Approach — UN proposal to identify irreducible floor of governance rules (human rights-based) rather than comprehensive global regimes
Contextualization — Adaptation of governance mechanisms to country/region-specific needs, priorities, and capacities
Bottom-Up Approach — User-needs and citizen-requirements-driven design (emphasized by Dr. Medi)

Methodology & Research Basis

The panel draws on:

UN policy development (Global Digital Compact, Scientific Panel, Global Dialogue mechanisms)
Country case studies — Ethiopia (digital ID), India (Aadhaar, UPI), Kenya (bilateral cooperation discussions)
Carnegie research — Jane Monga's Africa-focused research on AI development pathways and regional cooperation
Practitioner experience — Code Develop's funding and partnership work across DPI implementations

No peer-reviewed research papers, datasets, or quantitative studies are cited in the transcript; the discussion is policy-oriented and draws on institutional experience and ongoing negotiations.

Gaps & Limitations

Limited empirical data: No quantitative evidence on AI-DPI integration outcomes, adoption rates, or citizen benefit measurements
Africa-centric framing: While Jane Monga represents African perspectives, other Global South regions (Latin America, Southeast Asia) are underrepresented
Implementation detail: Discussions remain relatively high-level; granular technical governance solutions are not detailed
Risk assessment: Geopolitical disruption and fragmentation risks are acknowledged but not deeply analyzed with concrete scenarios
Funding solutions: Problems with current funding models are identified, but alternative sustainable models are not fully elaborated