African researchers push back against imported AI systems at inaugural law conference

JOHANNESBURG. Africa cannot afford to let its AI future be shaped by systems built elsewhere, on data drawn from elsewhere, governed by rules written elsewhere. That was the driving argument on the second day of the University of Johannesburg’s inaugural AI and the Law Conference, where researchers, legal scholars, technologists and industry leaders gathered to press on the foundational questions of AI deployment: who builds these systems, what data trains them, who profits, and who is held responsible when they fail.

The programme moved through keynote addresses, interdisciplinary panels and parallel research sessions. Dr Titus Mathe, Chief Executive Officer of the Technology Innovation Agency, contributed an industry perspective alongside academic and policy voices, positioning the conference as a cross-sector forum rather than a purely academic exercise.

United Nations Under-Secretary-General and Rector of the United Nations University, Professor Tshilidzi Marwala, opened by naming the questions that sit beneath nearly every consequential AI conversation. “When something goes wrong, or even when something goes right, who decides? Who is liable? Who designed the system that made the decision possible? And what trade-offs did we quietly accept along the way without really naming them?” he asked.

Marwala proposed that societies must evaluate AI through three interconnected dimensions: law, which determines liability and rights protection; governance, which shapes how oversight, data, algorithms and computing infrastructure are organised; and balance, which requires confronting unavoidable tensions between technological progress and its risks. “For every domain AI touches, society must decide how much certainty the law demands, how much structure governance provides and how much trade-off we are willing to accept,” he said.

The stakes are concrete. AI-powered legal assistance and document automation could substantially widen access to justice for communities historically excluded from legal services, but only if people can access these tools with confidence and if robust regulatory safeguards are in place. Marwala was direct about the risk of getting that wrong. “If we let the companies write the rules for the very tools meant to close the justice gaps, we risk the same digital divide simply reappearing inside the solution,” he cautioned.

Criminal justice applications carry particular hazards. Predictive policing and risk-scoring systems already influence bail, sentencing and parole decisions, yet they rest on statistical models that generate probabilistic outputs. Marwala questioned whether technologies built on probability should guide legal determinations that demand proof beyond reasonable doubt. He also warned that AI systems can produce responses that appear accurate and persuasive while being factually wrong, a limitation lawyers and policymakers must understand before they can regulate effectively. “Accuracy is not truth,” he said. “The same way we have long understood that correlation is not causation, we must recognise that accuracy is not truth.”

By contrast, University of Johannesburg Vice-Chancellor and Principal Professor Letlhokwa George Mpedi extended the argument into specifically African terrain. Transplanting governance frameworks designed for other regions and expecting them to address Africa’s distinct social, cultural and economic circumstances will not work, he argued. “AI models are mirrors. They reflect whatever they were fed. And if the data feeding those mirrors is overwhelmingly Western, then Africa is letting someone else’s mirror define its own reflection,” he said.

The risk reaches into African cultural heritage. Traditional fabrics, patterns and cultural expressions could be reproduced and commercialised by AI systems thousands of kilometres from their source communities, with no recognition, attribution or compensation. Mpedi pointed to South Africa’s use of geographical indication to protect Rooibos as a model, proposing that Africa develop comparable legal mechanisms to safeguard cultural provenance as AI systems proliferate. “Africa needs the AI equivalent of geographical indication, a legal mechanism that protects cultural provenance before it is absorbed and monetised elsewhere,” he said.

That leverage exists, and it is substantial. Through the African Union and the African Continental Free Trade Area, the continent encompasses 54 countries and approximately 1.4 billion people. “We should talk together. We represent 1.4 billion people. If we harness it, it will be strong,” Mpedi said.

Language is another operational gap. Africa hosts roughly 2,000 languages, yet leading AI systems function reliably in only a fraction of them. The implications for access to justice are direct. “An AI-driven legal aid tool is not neutral if it cannot reason competently in isiZulu or Sesotho,” Mpedi said. Algorithmic bias, in this framing, is not only about discriminatory outputs. It reflects whose languages, legal reasoning and social realities appear in the training data at all.

The two keynote addresses converged on a single structural argument. “Law alone arrives too late. It adjudicates harm only after the fact. Governance alone lacks teeth. It can design excellent systems that no one is actually bound to follow. And balance alone is just honesty about trade-offs, with no mechanisms to enforce the choices we have made,” Marwala said. “Put together, balance names the trade-offs, governance designs and manages them, and law enforces the outcome.”

The day continued with parallel sessions where scholars and practitioners presented work on AI, law and sustainable development. A panel on Artificial Intelligence, Sustainability and the Future of Society brought together Professor Georg Borges of Universität des Saarlandes in Germany, Professor Arthur Mutambara of UJ’s Institute for the Future of Knowledge, and Professor Sune von Solms of the UJ Faculty of Engineering and the Built Environment. The session was chaired by UJ Faculty of Law Acting Vice-Dean for Teaching and Learning Professor Sebo Tladi.

Mpedi closed with a remark delivered in the Kruger National Park, which marks a century of protection in 2026. He drew a parallel between conservation and the task now facing those building legal and governance frameworks for AI. “We are gathered in a place that protects something precious because generations before us had the foresight to build the legal instruments to make that protection durable. I hope this conference is one more such instrument,” he said. Whether the frameworks drafted here will carry that kind of durability remains the open question.