27/03/2026
We had the opportunity to present our policy brief, “Regulating AI-Driven Surveillance under the Protection of Personal Information Act (POPIA)”, at the 2nd African Cyber Law Conference, held at the Wits School of Law.
A key theme during the discussions was that fairness in AI begins with data. Without representative African datasets, AI systems risk reinforcing exclusion, making investment in local and multilingual data a key governance priority. In African contexts, fairness must also consider both individual and social groups outcomes to ensure equitable impacts across communities. Closely linked to this is the persistent gap between policymakers, regulators, and developers. Effective AI governance requires early and sustained collaboration across these groups, ensuring that systems are designed with both technical robustness and regulatory alignment in mind.
Discussions on transparency highlighted that explainability alone is insufficient. Instead, there is a need for systems that are not only technically interpretable but also accessible and understandable to the communities they affect. This aligns with emerging thinking around contextual and Ubuntu-informed AI frameworks, which emphasise designing technologies that reflect African values and lived realities rather than importing external governance models.
Another key theme was the shift from “responsible AI” to “resilient AI.” While principles such as fairness, accountability, and human rights remain essential, there is growing recognition that AI systems in Africa must also be adaptable and capable of operating effectively in complex, resource-constrained environments. This reinforces the importance of human-centred approaches, including participatory design and literacy-first strategies, to ensure communities have meaningful access to and agency over AI systems while mitigating risks such as algorithmic bias and entrenched inequalities.
The intersection of AI and intellectual property also generated considerable debate. Questions around authorship, ownership, and the use of copyrighted material in training AI systems remain unresolved. Many existing IP frameworks, largely shaped by global agreements, do not adequately reflect African interests or the realities of AI innovation. This raises important questions about whether these frameworks should be adapted, reinterpreted, or fundamentally reformed.
Key policy questions for the African context:
· Are we building responsible or resilient systems?
· Are our approaches proactive or reactive?
· And who sets the priorities?
Our takeaway:
AI governance in Africa is not only a legal or technical issue, but a broader socio-technical project that requires inclusive, context-sensitive, and forward-looking approaches.
