Federated Learning, Data Governance, and the Future of AI in Global Health: Reflections from the British Council Springboard Meeting
In May 2026, member of the AFRICAI-RI team travelled to Madrid to represent the project at the British Council Springboard Meeting on Fair and Responsible Use of AI in Global Health — a two-day gathering that brought together researchers, clinicians, industry representatives, and global health experts from across Europe and Africa. The meeting, organised by the London School of Hygiene and Tropical Medicine (LSHTM), focused on the opportunities and challenges of deploying artificial intelligence in low-resource healthcare settings. It was exactly the kind of interdisciplinary, cross-continental conversation that AFRICAI-RI was built for.
The Central Question: How Do We Make AI in Global Health Actually Work?
A recurring theme throughout the meeting was the risk of AI reinforcing, rather than reducing, existing inequities in healthcare. Disparities in imaging infrastructure, internet connectivity, maintenance capacity, and specialist availability all shape what AI tools can realistically do — and for whom. Speakers consistently emphasised the need for need-driven, not technology-driven AI development. Tools must align with local clinical priorities, infrastructure realities, and healthcare workflows. Technical performance alone is not enough.
AFRICAI-RI’s Contribution: Federated Learning as a Governance Model
We presented AFRICAI-RI’s approach within a dedicated session on federated learning as a model for ethical and collaborative data governance in multi-country AI projects. The presentation centred on a fundamental tension in international health AI research: building robust, generalisable AI models requires learning from diverse data, but patient data cannot and should not freely cross borders. Legal frameworks, sovereignty concerns, and community trust all place legitimate constraints on what can be shared.
- Imaging data remains securely stored within each partner institution
- Only model updates not patient data travel across the network
- Each site retains full ownership and control over its data
- Collaborative AI development proceeds without any centralised data pool
This approach strengthens privacy protection, supports local institutional ownership, and removes one of the most significant barriers to international AI collaboration: the requirement to centralise sensitive health data. The session also highlighted the broader governance dimensions of federated learning: participatory governance structures, stakeholder engagement, capacity building, and trustworthy AI principles, all of which are core to AFRICAI-RI’s design.
Key Themes from the Wider Meeting
Beyond the federated learning session, several themes resonated strongly with AFRICAI-RI’s work:
- Regulation is lagging behind practice. Many African countries still lack clear regulatory frameworks for AI in healthcare, which often forces institutions to rely on European or US certification systems. Addressing this gap is essential for responsible AI adoption at scale.
- Trust is non-negotiable. Research presented on how clinicians and healthcare workers in the Global South perceive AI underscored that transparency, explainability, and local stakeholder involvement are not optional extras, they are prerequisites for adoption.
- The technical and the clinical must be bridged. There was considerable interest in how AI systems can genuinely support clinical decision-making, from image quality control through to diagnostic support, in ways that are meaningful and usable in real-world settings.
Looking Ahead
The Springboard meeting reinforced that AFRICAI-RI is working at exactly the right intersection: federated infrastructure, ethical governance, local ownership, and clinical relevance. The conversations in Madrid will continue to inform how we shape the project’s approach to data governance, community engagement, and AI development as we move through Year 1 and beyond.
