India AI Impact Summit Insights
All sessions

South–South AI Cooperation: Building a Shared Policy Roadmap

Contents

Executive Summary

The Africa-Asia AI Policymaker Network, an over five-year collaborative initiative representing seven countries across Africa and Asia, convened to discuss how developing nations can build equitable, context-appropriate AI governance through peer learning and coordinated policy development. The session emphasized that effective AI policy requires holistic, socially grounded approaches driven by South-South cooperation rather than top-down adoption of Northern frameworks, with practical mechanisms including regulatory harmonization, shared infrastructure, talent mobility, and cross-border data governance.

Key Takeaways

  1. Peer learning > policy copying: Direct dialogue and knowledge exchange between policymakers from similar contexts is more valuable than importing finished regulatory documents. The Africa-Asia AI Policymaker Network's five-year success demonstrates this model's effectiveness.

  2. Infrastructure and data are collective challenges requiring collective solutions: No Southern country can build world-class AI infrastructure alone. Pooled GPU capacity, shared data platforms, and coordinated skill development represent practical pathways to bridging the compute and talent gaps.

  3. Trust-building is prerequisite to governance: Before harmonizing regulations or sharing data, Southern countries must establish institutional confidence mechanisms. Cross-border data frameworks, shared audit standards, and demonstrated good governance build the trust necessary for cooperation.

  4. Context-specific AI (language, culture, sector focus) is non-negotiable: Sovereign models, local language support, and sector-specific solutions (agriculture, health, informal workforce) are not luxury features but prerequisites for AI to serve Southern populations. This requires deliberate policy choices and resource allocation.

  5. Implementation requires systems thinking: Successful AI deployment in Southern contexts depends on multiple aligned interventions: compute infrastructure, data governance, talent development, regulatory clarity, and sectoral coordination. Treating these as isolated challenges perpetuates the scaling gap.

Key Topics Covered

  • South-South AI cooperation frameworks — mechanisms for policy exchange, knowledge transfer, and coordinated governance among Global South countries
  • Regional AI policy development — national AI strategies and governance approaches across Ghana, India, Indonesia, Kenya, Rwanda, South Africa, and Uganda
  • Infrastructure and compute challenges — GPU allocation, shared compute capacity, and the digital divide in AI development
  • Data governance and sovereignty — frameworks for cross-border data sharing, DPI (Digital Public Infrastructure), and sovereign data strategies
  • Talent development and skills — AI literacy programs, capacity building, workforce development, and microcredential recognition
  • Ethical and responsible AI — governance, bias mitigation, safety institutes, and alignment with local cultural values
  • Regulatory sandboxes and experimentation — innovation-friendly governance models with appropriate safeguards
  • Sectoral priorities — agriculture, health, education, and public service delivery as focus areas for AI deployment
  • Trust and accountability — data sharing principles, cross-border collaboration frameworks, and institutional mechanisms
  • Global platforms and coordination — Africa AI summit, regional strategies, and continental governance networks

Key Points & Insights

  1. Policy learning requires dialogue, not documents: Regulatory frameworks alone cannot transfer knowledge. Policymakers need direct exchange to understand how decisions were made, which choices are context-specific, and how stakeholder consultation shaped outcomes — this is the core premise of the Africa-Asia AI Policymaker Network.

  2. Infrastructure pooling is economically viable: Rather than each country building isolated AI infrastructure, Southern countries can collectively develop shared compute resources (GPUs, cloud platforms) at lower cost — India's model of providing 10,000 GPUs at subsidized rates (~$1/hour) demonstrates this approach.

  3. Sovereign AI models are essential for linguistic and cultural diversity: India's development of language-specific LLMs (supporting 22 Indian languages) and voice-based AI reflects the recognition that Western LLMs don't serve diverse populations; Southern countries must develop their own models aligned with local contexts.

  4. Trust is the foundational challenge in cross-border data sharing: Multiple speakers identified trust as the critical barrier to data exchange and cooperation. Without institutional frameworks and demonstrable confidence, countries revert to data silos despite acknowledging the benefits of sharing health, agricultural, and weather data.

  5. Implementation scaling remains a critical gap: Rwanda and other countries report success in developing AI pilots and MVPs but struggle with scaling to real-world impact. The AI Scaling Hub model (providing compute resources, data access, and regulatory support) addresses this by treating scaling as a systemic challenge requiring multiple interventions.

  6. Multistakeholder engagement must begin at cabinet level: Ghana's experience shows that AI policy documents stall without early engagement from cabinet ministers. Including decision-makers and AI focal persons across ministries prevents implementation bottlenecks and ensures adoption.

  7. DPI (Digital Public Infrastructure) enables cost-effective, scalable governance: India's success with systems like UPI, Jan Dhan, and national ID stacks demonstrates that AI governance and deployment can leverage existing digital infrastructure rather than requiring parallel systems — this model is transferable to other Southern countries.

  8. Ethical AI governance is an enabler, not a barrier: Multiple speakers emphasized that ethics frameworks (addressing bias, transparency, safety) accelerate rather than impede innovation by building public trust and regulatory credibility. Indonesia's framing of ethics as a "guardrail" rather than obstacle is instructive.

  9. Regulatory harmonization and mutual recognition reduce fragmentation: Aligning AI governance across regions (e.g., East African Community strategy alongside national strategies) prevents regulatory arbitrage and allows innovation zones (sandboxes) to function as learning platforms where governments observe success and failures collectively.

  10. South-South cooperation produces tangible, documented outputs: The network has generated an AI policy playbook, Africa AI Declaration, Africa AI Council, and use case repositories — concrete tools that reduce policy development time and provide ready-made solutions for new initiatives (e.g., Kenya implementing soil analysis solutions from Telangana).

Notable Quotes or Statements

  • Wolfgar Bungart (German Ministry for Economic Cooperation): "These documents, they cannot talk to you... If these documents cannot talk to you, you should talk to each other." — Capturing the rationale for the peer-learning network model.

  • Odilla Aayodelli (HSRC, South Africa): "AI can't be treated as purely a technical project and governance can't be left as a black box... we need to think holistically about how to build responsive AI policy and systems."

  • Katarina Adam (Indonesia): "The heart of AI is not the code, it's the people... let's write [the future of AI] together" and "the story of AI is often told as a race between superpowers. But looking at this panel, I see a different story."

  • Rama Devi Lanka (NITI Aog, India): "We have 490 million informal workers... carpenters, construction workers, electricians... they contribute more than 50% to India's GDP... can we build a credential system and tools that increase their productivity?" — Highlighting AI's potential for informal economies.

  • Cyprien Shamiima (Rwanda): "We learned that Rwanda was not going to do it alone... South-South collaboration is quite important" — Framing collaboration as strategic necessity, not charity.

  • Maxwell Ababio (Ghana, Data Protection Commission): "Trust is actually the issue... once you have trust, sharing information becomes easy. If not, we all live in a 'truth economy' where we don't trust any information we share." — On the foundational challenge of cross-border cooperation.

Speakers & Organizations Mentioned

Government & Policy Leaders

  • Rachel Adams — CEO, Global Center on AI Governance (moderator)
  • Wolfgar Bungart — Senior Policy Officer, German Federal Ministry for Economic Cooperation and Development (GIZ)
  • Odilla Aayodelli — Senior Research Specialist, Human Sciences Research Council (HSRC), South Africa
  • Richard Kiari — Deputy Director of ICT, National Treasury, Government of Kenya
  • Katarina Adam — Coordinator of Data Protection and Information Security Management, Ministry of Communication and Information Technology, Government of Indonesia
  • Cyprien Shamiima — Program Manager, AI Scaling Hub, Center for the Fourth Industrial Revolution, Rwanda
  • Rama Devi Lanka — Lead, Frontier Tech Hub for State-led Engagement, NITI Aog, Government of India
  • Irene Karungi Seito Leco — Senior ICT Infrastructure Engineer, Ministry of ICT and National Guidance, Government of Rwanda
  • Azazi Vuvuzela — Director of Cyber Security, Department of Communications and Digital Technologies, South Africa
  • Maxwell Ababio — Head of Technology and Ethics, Data Protection Commission, Ghana

Supporting Organizations

  • Global Center on AI Governance (hosts, policy research)
  • GIZ (German Society for International Cooperation) — funds network infrastructure
  • Human Sciences Research Council (HSRC), South Africa — provides evidence base and policy support
  • Center for the Fourth Industrial Revolution (C4IR), Rwanda — implements AI scaling initiatives
  • NITI Aog (National Institution for Transforming India) — coordinates India's AI mission and frontier tech strategy
  • Smart Africa — leads continental AI governance initiatives
  • Gates Foundation — funds AI Scaling Hub initiatives
  • UNESCO — supports ethical AI readiness assessments
  • Microsoft — provides AI infrastructure partnerships (South Africa commitment: 300 million rand)
  • UAE (Dubai Center for AI) — platform for South-South talent/certification initiatives

Peer Network Members (7 countries)

Ghana, India, Indonesia, Kenya, Rwanda, South Africa, Uganda

Technical Concepts & Resources

Infrastructure & Compute

  • GPU allocation programs — India's National AI Mission distributes 10,000 GPUs at subsidized rates (~60 rupees/hour, <$1)
  • Shared compute capacity — collaborative approach to pooling GPU and cloud resources across Southern countries
  • Microsoft AI infrastructure partnership — 300 million rand commitment (South Africa)
  • G42/UAE partnership — compute and AI system partnerships available to participating countries

Data & Models

  • Sovereign AI models — country-specific large language models (e.g., Bhaaratiya, Indic LLMs in India supporting 22 languages)
  • Voice-based AI — language-localized voice interfaces for low-literacy populations (agricultural advisory, public services)
  • National data platforms — centralized repositories (India example: 2,000+ datasets available for startups, universities)
  • Multilingual datasets — development of local-context datasets reflecting regional languages and cultural contexts
  • DPI (Digital Public Infrastructure) — foundational systems (UPI, Jan Dhan, national ID stacks) enabling scaled AI deployment

Governance & Frameworks

  • AI policy playbook — practical guidance document produced by the network for regulators
  • Africa AI Declaration — adopted continental governance framework
  • AI ethics guidelines — Indonesia's updated national ethics framework; UNESCO ethical AI readiness assessments
  • Regulatory sandboxes — controlled experimentation zones for testing AI systems under government supervision
  • Cross-border data frameworks — principles for harmonized data sharing (African community framework)
  • AI focal persons model — designated officials in each ministry coordinating AI adoption and oversight (Ghana model)

Capacity & Skills

  • AI talent factory — Indonesia's project-based learning model bridging education-industry gap
  • Microcredential recognition — portable digital certifications across borders (proposed)
  • AI fluency courses — Rwanda's mandate for all civil servants; Dubai Center for AI platform serving 100,000+ officials
  • AI fellowship programs — Rwanda-Singapore-UAE collaborative learning initiative
  • Prompt engineering training — chief directors and ministry staff training (Ghana model)

Use Cases & Solutions Repositories

  • NITI Aog use case repository — documented AI solutions across agriculture, education, health sectors (available on website)
  • AI scaling pilots — health, agriculture, education sector-specific implementations (Rwanda focus)
  • Soil analysis solution — Telangana project successfully implemented in Kenya
  • Informal workforce productivity tools — AI for carpenters, electricians, construction workers (India roadmap)

Assessments & Planning Documents

  • National AI landscape assessments — Uganda's comprehensive ecosystem mapping
  • Economic impact studies — Rwanda pre-generative AI estimate: 6% GDP contribution (noted as conservative)
  • AI for economic development roadmap — India's framework identifying 10x productivity gains in banking, manufacturing, automobiles
  • AI for informal workforce roadmap — India's strategy for 490 million workers in unstructured economy
  • Digital transformation roadmaps — Uganda's 2024 roadmap positioning AI as national priority

Note on Accuracy: This summary is based entirely on the provided transcript. No claims are presented beyond what speakers articulated. Where specific timelines, numbers, or project names are mentioned, they are transcribed as stated; some audio quality variations mean minor transcription uncertainties in proper names or exact figures are noted where relevant.