Unveiling the AI Governance Chasm at Wits University

The University of the Witwatersrand (Wits) in Johannesburg has emerged as a pivotal voice in addressing a critical issue in artificial intelligence (AI): the profound disconnect between global AI systems and African users. On April 15, 2026, Wits published insights from its second African Cyber Law Conference, held March 24-25, 2026, under the theme 'Resilient and Responsible Design: Governing AI, expression and digital media.' Organised by Dr Nomalanga Mashinini, senior lecturer at the Wits School of Law, the event gathered legal scholars, students, and digital professionals to dissect how AI, largely designed without African input, perpetuates exclusion and harm.

Dr Mashinini explained that AI's foundational datasets, predominantly from the Global North, fail to capture African linguistic diversity and cultural contexts. Harmful content in local languages often evades detection, rendering entire populations invisible or misclassified. This 'systemic exclusion,' as she termed it, manifests in everyday applications like content moderation on social platforms, where nuances in Zulu, Swahili, or isiXhosa are lost, amplifying misinformation and bias.

Such revelations underscore a governance chasm—not a dearth of laws, but uncoordinated enforcement across Africa's fragmented regulatory landscape. South Africa's universities, including Wits, are at the forefront, leveraging research and policy advocacy to bridge this gap.

AI Design Biases: Why Africans Are Sidelined

AI systems exhibit biases rooted in their training data. Facial recognition technologies, for instance, perform poorly on darker skin tones prevalent across Africa, with error rates up to 34% higher for Black faces compared to lighter ones, according to studies on global models. Language models fare worse: only 2% of training data for major AI comes from Africa, leading to subpar handling of over 2,000 African languages.

In South Africa, this translates to real-world issues in higher education. Automated grading tools misinterpret student submissions in indigenous languages, while recruitment AI in universities favors Western resumes, disadvantaging local talent. Wits researchers highlight 'digital extractivism,' where African user data is harvested by foreign tech giants, processed abroad, and monetized without local benefits, exacerbating inequality.

Professor Jonathan Klaaren from Wits School of Law emphasized the need for alignment between legal frameworks, technical systems, and societal needs. Existing South African laws—like the Protection of Personal Information Act (POPIA) and cybercrime legislation—provide a foundation, but siloed institutions hinder effective oversight.

Wits MIND Institute: Forging African-Led AI Innovation

Wits' Machine Intelligence and Neural Discovery (MIND) Institute stands as Africa's premier hub for interdisciplinary AI research, launched to converge natural and artificial intelligence. Funded by Google.org ($1M in 2025) and others, it prioritizes AI policy and governance tailored to African contexts, tackling data privacy, cybersecurity, and sustainable development.

The institute trains PhD candidates and postdocs, drawing diaspora talent, and collaborates on large-scale models for drug discovery and astronomy. Hosting the African Machine Learning Indaba (AMLD Africa 2026) in January, Wits fosters continent-wide talent, with events drawing thousands to discuss AI for healthcare and agriculture.

Through the AI Africa Consortium, Wits links researchers with global partners like IBM, ensuring African voices shape ethical AI deployment.

South Africa's National AI Policy: A Framework for Universities

The Draft National Artificial Intelligence Policy Framework, gazetted April 10, 2026, positions higher education as central to AI advancement. It mandates AI integration in curricula from primary to tertiary levels, educator training, and ethical focus on biases and societal impacts. Universities like Wits and UCT are tasked with research hubs, talent pipelines, and industry partnerships.Learn more about the policy

Proposals include a National AI Commission, Ethics Board, and Regulatory Authority, with risk-based oversight. For higher ed, this means sandboxes for AI tools in teaching, bias audits, and incentives for startups addressing local challenges like multilingual education.

Other South African Universities Pioneer AI Ethics

North-West University (NWU) led with South Africa's first official AI policy in January 2026, approved by council. It promotes human-centered use in teaching, research, and assessments, with an AI Steering Committee overseeing ethics and sustainability concerns like energy consumption.Read NWU's policy details

University of Cape Town's (UCT) AI Initiative reimagines ethical AI for African futures, focusing on health, climate, and inequality through foundational research. Stellenbosch University issued a 2025 position statement on ethical AI in research, mandating transparency and bias mitigation. University of Johannesburg (UJ) co-hosts the National AI Institute, advancing applied AI in key sectors.

These efforts respond to rising GenAI use: 92% of African HE students now employ AI tools, up from 66% in 2024, but only 40% of institutions have policies, per recent surveys.

Challenges in South African Higher Education

AI adoption surges—62% via learning management systems—but governance lags. Academic integrity cases rise, with undetected GenAI in assignments. Biases affect diverse students: language models falter on African tongues, disadvantaging non-English speakers comprising 80% of SA population.

• Risk of job displacement for lecturers without upskilling.
• Data privacy breaches in AI-driven analytics.
• Infrastructure gaps: limited bandwidth hampers AI tools.
• Equity issues: historically disadvantaged institutions trail in funding.

UKZN's progressive guidelines encourage AI augmentation while enforcing citation, mirroring global trends but adapted locally.

Case Studies: AI Bias and Responses in SA Unis

At Wits, facial recognition pilots for campus security revealed 25% higher errors for Black students, prompting algorithmic audits. UCT's health AI projects mitigate biases in diagnostic tools trained on diverse SA datasets.

Stellenbosch's framework requires human oversight in AI research, reducing bias risks.

Solutions: Coordinated, Context-Specific Governance

SA universities advocate embedding ethics in design: transparency, accountability, impact assessments. Wits calls for pan-African coordination, leveraging POPIA for data sovereignty.

Photo by Nel Ranoko on Unsplash

• Multilingual datasets: Initiatives to train models on African languages.
• Talent programs: PhDs, fellowships via MIND and UCT.
• Policy briefs from Wits conference target surveillance, fairness.
• Industry ties: Google, IBM partnerships for infrastructure.

Future Outlook: SA Higher Ed Shapes African AI

With the National AI Policy advancing, universities like Wits position SA as Africa's AI leader. Projections show AI adding 15.7% to GDP by 2030 if governed well. Challenges persist, but proactive policies—NWU's pioneering framework, UCT's ethical focus—equip students for an AI-driven world.

Stakeholders urge investment in infrastructure and training. As Dr Mashinini notes, 'Conversations must continue' for implementation. SA higher ed's response offers a blueprint: inclusive, resilient AI governance for all Africans.

For academics and students, opportunities abound in research jobs and ethical AI careers. Explore positions at leading SA institutions to contribute to this vital field.