Unisa law candidate puts ubuntu at the centre of South Africa's AI rules debate

A Unisa law candidate has put ubuntu at the centre of South Africa’s artificial intelligence debate, arguing that the country’s AI systems should be judged by whether they advance human dignity rather than by how efficiently they run.

Thabiso Bob Mbuyisa, a public-interest writer, made the case in a thought-leader piece published by the Mail & Guardian on 16 September 2026. His argument is straightforward: AI is not a neutral tool, and South Africa has a choice about the values it builds into the systems it adopts.

“Ubuntu is often summarised by the phrase umuntu ngumuntu ngabantu, a person is a person through other people,” he writes, describing it as an ethical philosophy of shared humanity and mutual responsibility rather than a cultural slogan.

Where AI is already touching South African lives

Mbuyisa lists healthcare, education, financial services, public administration and the justice system as areas where AI is beginning to influence how decisions get made.

That is not a distant prospect. South African banks use automated systems to score loan applications. Hospitals triage patients with software support. Government departments are digitising records and, in some cases, automating eligibility checks. Each of those systems makes choices about people, and each of those choices is built on data that reflects the country’s past.

Mbuyisa draws on the work of scholars Safiya Umoja Noble and Kate Crawford to make the point that algorithms learn from historical data. Where that data carries discrimination or inequality, the systems can reproduce and amplify it rather than correct it.

For South Africa, that is not an abstract concern. Apartheid-era spatial planning, credit records and employment patterns are still embedded in the datasets that modern systems train on. A hiring algorithm trained on decades of skewed workplace data will not produce a fair shortlist just because nobody typed a racial preference into the code.

The dignity argument

Mbuyisa’s central claim is that AI systems classify people by measurable characteristics and predictive models, which risks reducing human beings to datasets. He argues that ubuntu holds that every person has inherent dignity that cannot be reduced to a probability score.

He makes the point sharply on gender justice. AI trained on biased historical data, he writes, may reproduce discriminatory hiring practices, unequal healthcare outcomes or unfair policing patterns. Those are not technical glitches. They are ethical failures.

This is where the piece is at its most useful for a South African reader. The country already has a constitutional framework that treats dignity, equality and non-discrimination as binding commitments. Mbuyisa is effectively asking whether AI governance will be held to the same standard.

What he says responsible AI requires

Mbuyisa sets out four conditions: fairness, transparency, accountability and meaningful human oversight. He argues these are not obstacles to innovation but the conditions that make innovation socially legitimate.

He also calls for collaboration between government, universities, industry and civil society, and for communities affected by AI systems to have a meaningful voice in how those systems are designed, implemented and evaluated.

That last point matters in practice. South Africa has a relatively thin AI governance landscape compared with the European Union, which passed its AI Act in 2024. The country does not yet have dedicated AI legislation. Work on a national AI policy framework has been underway through the Department of Communications and Digital Technologies, and the Presidential Commission on the Fourth Industrial Revolution has produced recommendations on AI and data governance.

The gap Mbuyisa identifies is not that South Africa lacks ethical language. It is that the language has not yet been translated into enforceable rules for the systems already in use.

Where the piece stops short

Mbuyisa’s argument is coherent and it is grounded in scholarship he names. What the piece does not do is engage with the specific South African institutions that would have to carry the framework he proposes.

There is no mention of the Information Regulator, which enforces the Protection of Personal Information Act and has been the country’s most active data governance body. There is no reference to the Competition Commission’s online intermediation market inquiry, which has already examined how platform algorithms affect South African businesses. There is no discussion of the Department of Communications and Digital Technologies’ policy process or the timeline for any AI-specific regulation.

Those are the levers that would turn ubuntu from a principle into a rule. A reader who finishes the piece knowing what should guide AI, but not who would enforce it or by when, has half the story.

The piece is also silent on cost. Human oversight, transparency requirements and impact assessments all carry a price. For a South African startup building a credit model or a township clinic deploying a triage tool, those costs are real and they shape whether ethical AI remains the preserve of well-funded institutions.

The argument that holds

None of that undermines the core claim. Mbuyisa is right that AI is a social and political technology, not just a technical one. He is right that the choices embedded in algorithms reflect whose voices mattered when the system was built. And he is right that South Africa has an opportunity to develop a human-centred approach that other countries could learn from.

The country has done this before. South Africa’s constitution was drafted with an explicit commitment to human dignity that was unusual at the time. The same instinct, applied to AI, would be consistent with that tradition rather than a departure from it.

Mbuyisa’s closing argument is that the question is not whether AI will shape South Africa’s future, but whether that future will be rooted in ubuntu. It is a fair question. The harder one, and the one his piece leaves open, is who will answer it and by when.

Source: Mail & Guardian, Ubuntu must guide artificial intelligence before artificial intelligence guides us, 16 September 2026.