South Africa: UCT Research Urges Urgent AI Regulation for Human Rights

Nokuthula Olorunju, University of Cape Town PhD graduand and researcher. South Africa needs to move beyond its longstanding “wait-and-see” approach to artificial intelligence (AI) regulation and establish a binding, constitutional framework that protects human rights without stifling technological innovation. This is the argument advanced by University of Cape Town (UCT) PhD graduand Nokuthula Olorunju, whose thesis examines how South Africa should regulate AI in the face of growing adoption and increasingly complex risks. Olorunju will be capped with a PhD in Law (Commercial Law) on 8 September. Her thesis is titled: “Artificial intelligence, bias and the rule of law: A constitutional approach to the regulation of AI in South Africa”. While SA has previously faced more pressing regulatory priorities, Olorunju argues that the rapid growth of AI means the country can no longer delay putting an effective regulatory framework in place. The research comes as SA grapples with the development of its national AI policy . In April, communications and digital technologies minister Solly Malatsi withdrew the country's draft National AI Policy after internal checks confirmed its reference list contained fictitious sources, saying the failure had compromised the document's integrity and credibility. Olorunju's research uses theoretical, doctrinal and comparative methodologies, examining AI regulation through the lenses of law and technology, human rights, constitutional law and international law. Her research concludes there is no perfect model for regulating AI. Instead, countries should begin with frameworks they can realistically implement, taking into account their regulatory constraints. For SA, she identifies fragmentation, weak implementation and longstanding legal silos as factors that can create opportunities for AI misuse. Legislative fragmentation can lead to overlapping jurisdictions, uncertainty around authorisation processes, conflicting mandates and confusion over which institutions are responsible for particular aspects of AI governance. “In essence, if a gap can be exploited, it most probably will be,” she said. Olorunju proposes a constitutional approach as a hybrid framework aimed at closing these gaps, while addressing practical enforcement and implementation challenges. The approach would include clarifying conflicting mandates, creating an interdisciplinary AI oversight body and avoiding unnecessary duplication of governance structures. Olorunju argues that existing legislation, non-binding ethical principles and self-regulation are not sufficient to address the risks associated with AI. “While ethical principles and self-regulation have helped address some regulatory gaps, they do not carry the force of law. Self-regulation can also result in organisations selecting principles that are easier to incorporate into their existing missions and values. Government therefore has a responsibility to ensure enforceable laws are in place to address grey areas and loopholes that could allow rights to be exploited,” she argues. The complexity of AI also means regulation cannot be developed in isolation. Olorunju advocates an interdisciplinary approach that brings together experts from multiple fields to inform legislation and assist with its implementation. Human rights are central to Olorunju's proposed approach, particularly because AI systems can reproduce and potentially amplify existing inequalities. Her research highlights risks, including AI-enhanced cyber crime, data breaches, privacy violations involving deepfakes, AI systems that are not trained on South African data, racial and gender bias in lending and employment, copyright infringement and mass surveillance. According to the study, AI systems can reproduce historical inequalities when the data used to train them reflects existing societal biases. “Given South Africa’s history, it’s the perpetuation of racial and gender bias, the exponential increase of the digital