Memory is infrastructure — the dangerous cost of dissolving the National Arts Council board

Every few years, South Africa finds itself trapped in the same conversation.

Foreign nationals are blamed for unemployment. Communities turn on migrants from elsewhere on the continent. Politicians offer increasingly simplistic explanations for complex social and economic challenges. The names, instigators and locations change. But the underlying question remains remarkably consistent.

Who belongs here?

The usual explanations are familiar. There are not enough jobs. Public services are under pressure. The government has failed. These explanations are not wrong. They are incomplete. Because beneath the anger sits something older. A society still trying to understand itself.

We have inherited competing stories about who this country belongs to and how it came into being. A country where some still imagine civilisation arrived with colonial settlement, while others point to millennia of human presence, trade, governance and cultural production that existed long before 1652. The result is that South Africans often find themselves arguing about the future of the country through disputes about the past.

Who arrived first? Who belongs? Who has a claim? Who gets to benefit?

These are not merely historical questions. They are democratic questions.

South Africa occupies a peculiar place in history. Internationally, and within historical records online, it is understood as one of Africa’s older sovereign states because the Union of South Africa was established in 1910, decades before the wave of decolonisation that swept the continent in the 1950s and 1960s. Yet for the majority of South Africans, freedom did not arrive in 1910. It arrived in 1994. This means South Africa is simultaneously one of Africa’s older states and one of its youngest democracies.

The contradiction matters.

Because democracy is not simply about elections or constitutions. It is also about a society’s ability to tell coherent stories about itself. Who we are. How we got here. What we owe one another. And what kind of future we are trying to build together. This is where memory becomes infrastructure.

Memory as infrastructure

When we think about infrastructure, we usually think about what holds a country together physically: roads, ports, power stations, railways, the like. We rarely think about memory. Yet memory is also infrastructure. History books, museums, libraries, archives. All of these are infrastructure.

What these institutions provide is not simply information. They provide continuity. They help society connect past decisions to present realities. They preserve the stories through which people make sense of themselves and one another. Without them, public debate becomes detached from historical understanding. Identity becomes grievances. History becomes a rumour.

Much of my work within AI ethics has been shaped by a deceptively simple concept: legibility. To be legible is to be visible to a system of power. To be counted, recorded, recognised, represented. It is the difference between existing and being seen to exist.

Every system of governance depends on legibility. States rely on maps, censuses, records and archives to make populations visible to themselves. What gets recorded shapes what gets resourced. What gets classified shapes what gets protected. What disappears from the archive disappears, eventually, from policy, memory, from public life. Democracy, in particular, depends on institutions capable of preserving a shared and contested account of reality, because without that account, there is no common ground from which to argue, organise or govern.

Colonialism and apartheid understood this. They were not only systems of territorial and economic control. They were systems of legibility, deliberately designed to determine whose histories were recorded, whose languages were preserved, whose knowledge counted and whose experiences were considered worthy of documentation. Entire communities were rendered invisible through administrative and epistemic processes that privileged some stories and suppressed others. The archive was never neutral. It was a political instrument.

What is less understood is that AI systems operate on exactly the same logic. They can only learn from what has been preserved. They can only recognise what has been documented. They can only represent what is present in the data. Legibility, in other words, is no longer only a question of who the state can see. It is also a question of who AI can see. And the two are connected, because the archives that AI systems learn from were built by the same institutions, under the same political pressures, that shaped what the state chose to record in the first place.

Public memory in the age of AI

This is what makes the dissolution of the National Arts Council (NAC) board more than an administrative matter. An institution that supports cultural production helps determine what enters the archive. Which stories get told. Which languages are documented? Which communities become visible. In weakening it, we are not simply cutting arts funding. We are narrowing what South Africa will be able to remember about itself, and what future AI systems will be able to learn.

The NAC was established in 1997 as part of South Africa’s broader democratic reconstruction project. Its mandate was never simply to fund artists. It was to support literature, theatre, music, dance, indigenous cultural expression and community arts initiatives through which South Africans could begin telling fuller and more inclusive stories about themselves. In many ways, institutions like the NAC form part of the democratic infrastructure through which South Africa can remember itself and begin to construct a unified identity.

Which is why what is happening to it right now should concern all of us.

The recent dissolution of the NAC board by Minister of Sports, Arts, and Culture Gayton McKenzie has been treated, largely, as an administrative matter. The public debate has understandably focused on governance failures, labour disputes and institutional dysfunction. Public institutions should be accountable. Where reform is necessary, reform should happen. But accountability should not obscure a larger question: what exactly is being weakened, and at what moment?

The decision arrives while South Africa is visibly struggling with the question of belonging, identity and representation that institutions like the NAC were established, in part, to help societies navigate. So it is evident we are not dissolving this council in a vacuum. We are dissolving it while the wounds it was built to help heal are still open.

And there is a further dimension to this that has received almost no attention, one that will define the stakes not just for this decade but for the next century.

AI introduces a new and urgent dimension to the challenge of legibility. AI systems learn through datasets, archives, images, books, recordings and digital traces. Like states, they can only act upon what they can see. And they cannot see effectively and disappear from view.

This is why cultural institutions now matter in ways we have not fully appreciated. They do not simply preserve memory for people. Increasingly, they preserve memory for AI.

In the twentieth century, archives helped societies remember. In the twenty-first century, they help determine how AI systems understand the world. AI systems cannot learn from knowledge that has not been preserved. They cannot recognise histories that have not been documented. They cannot accurately represent communities that remain absent from the archive.

South Africa already exists within a global digital ecosystem where representation is profoundly uneven. Most digitised knowledge originates from Europe and North America. African languages remain significantly underrepresented. Oral traditions, community-based knowledge systems and local cultural histories are poorly documented compared with their Western counterparts.

If cultural institutions continue to weaken while AI systems are increasingly trained on knowledge produced elsewhere, we risk reproducing old exclusions through new technologies. The colonial project was, in part, a project of erasure. The democratic project has been, in part, a project of recovery. The danger is that we mistake administrative failure for institutional irrelevance, and in doing so, allow that recovery to stall at precisely the wrong moment.

What we are witnessing in the tensions around migration, belonging and identity is not simply a failure of economic and migration policy. It is also what happens when a society loses the institutions responsible for helping it understand itself.

South Africa’s future sovereignty will not depend only on who owns the algorithms, the data centres or the compute infrastructure. It will also depend on whether we preserve the stories, languages, histories and cultural knowledge from which those systems learn what this society is. Reform the NAC where reform is necessary. Hold it accountable. Demand that it fulfils its mandate with rigour and transparency. But do not mistake dysfunction for disposability.

What a society forgets, AI will never know. And what machines never learn may become increasingly difficult for future generations to recover. DM