As AI governance conversations across Africa mature, it is becoming increasingly important to look beyond laws, policies and strategies and understand the AI systems those frameworks are trying to govern.
AI models are becoming more capable and are being built into digital products, public services, businesses, education, research, financial services and other parts of society. Yet many AI governance conversations still discuss “AI” in general terms, without looking closely at the models that sit underneath these systems.
This matters because a relatively small number of foundation models can be adapted for many different uses. Their capabilities, limitations, training data, safeguards and access conditions can therefore have consequences far beyond the original model. This is the starting point for the Africa Foundation Model Project.
The project will build an Africa-focused evidence base on the foundation models increasingly shaping the AI ecosystem. Our work will look at:
-
Models: Who is developing them and what capabilities are they acquiring?
-
Data: What do we know about the data behind them, including the representation of African languages and contexts?
-
Access: Who can build, study and use these models, and what role do open-weight models play?
-
Disclosures: What do developers tell us about safety, evaluation, limitations and model behaviour?
-
Deployment: Where and how are these models being used or adapted in Africa?
-
Governance: What questions do these developments raise for African policymakers, regulators, businesses, researchers and civil society?
We will begin with a grounding series that explains the basic concepts behind foundation models from emergence and homogenisation to compute, model documentation, open weights, capabilities and safety.
From there, the project will track important developments and analyse what they mean for Africa. This project complements, rather than replaces, our monthly AI governance newsletter. The newsletter will continue to provide a broad view of AI governance developments across Africa and global developments that are relevant for Africa. The Foundation Model Project takes a different lens by looking at the systems themselves.
Our approach is deliberately focused. We are not building models or attempting to technically benchmark every system. Instead, we will use publicly available evidence to understand what is known, what developers disclose, what remains unclear, and why it matters.
Our goal is to help African stakeholders better understand the AI systems increasingly shaping the continent's digital future.
As foundation models become more powerful and widely embedded in society, understanding the technology itself is becoming part of understanding its governance. The Africa Foundation Model Project is our contribution to that work.
