Fifty Years After Soweto, Can We Still Trust Historical Photographs in the AI Age?
In 1976, photographers risked everything to document the uprising. Now AI can fake history in seconds.

Sign up for my Substack daily AI newsletter here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
On 16 June 1976, Sam Nzima photographed a dying Hector Pieterson being carried through the streets of Soweto. Photographers who documented the uprising were beaten and arrested, and some smuggled their negatives past the police. Peter Magubane, one of them, put it simply: “A struggle without documentation is not a struggle.” Fifty years later, that documentation faces a threat its creators could not have imagined. Generative AI can now produce convincing “historical” photographs of events that never happened, in seconds. In the anniversary year of Soweto, South Africans should ask what this means for the most powerful evidence of our past.
CONTEXT AND BACKGROUND
The problem is already well documented elsewhere. In January, Germany’s concentration camp memorials, including Bergen-Belsen, Buchenwald and Dachau, warned that AI-generated images of invented scenes distort history and erode trust in authentic documents. They called on platforms to label such content and stop it being monetised.
In Australia, spam pages have posted AI-generated “vintage” photographs of real Holocaust survivors alongside invented life stories. One survivor said simply that the image “is not of me”. The Auschwitz Memorial warned that such posts “replace authentic historical evidence with fiction.”
INSIGHT AND ANALYSIS
The most immediate damage is false images. The more serious damage is doubt. Historian Clara Mansfeld of the Hamburg memorials asks what happens “when the first thought we have with every historical image is, ‘Is that even real or is that AI?’” She notes that some accounts profited from fake Holocaust imagery, and platforms acted only after journalists exposed them. South Africa’s own reckoning depended on evidence. The Truth and Reconciliation Commission heard from about twenty-one thousand victims as it set out to reveal the truth about serious abuses committed between 1960 and 1994. Where some still minimise apartheid’s brutality, cynicism about genuine photographs and records could become a political weapon.
Technology offers little reassurance. When a fabricated image of a 9/11 hero went viral this year, leading AI chatbots gave mixed or wrong answers about whether it was real. About a quarter of Americans alive today were born after that day, so what they see online shapes what they believe happened. Almost all of today’s South African learners were born after apartheid ended.
I have previously written about this in the context of deepfake fraud. I argued that cheap, convincing fakes make sight and sound unreliable as proof, so people and organisations must replace instinctive trust with deliberate verification habits. The same principle now applies to our history.
South Africa is not immune. This month, Africa Check debunked an AI-generated graphic falsely claiming that all Home Affairs offices had shut down. It is a short step from fake notices about the present to fake images of the past.
IMPLICATIONS
Museums and archives hold the originals, which makes them the anchors of trust. They should make authentic collections easier to find online, publish clear provenance for iconic images and be ready to respond quickly when fakes circulate. Platforms should label AI-generated historical imagery and stop rewarding it with revenue.
Educators carry the longest-term responsibility. Learners should be taught to ask where a historical image comes from, who took it and which archive holds it. Journalists should treat any unfamiliar “historical” photograph with the same scrutiny they apply to breaking news.
CLOSING TAKEAWAY
The photographers of 1976 understood that evidence is power. Their images helped the world see what apartheid was, and they paid a price to make them. AI does not erase those photographs, but it can surround them with plausible fiction until the real and the invented look the same. The answer is not to distrust every image, but to invest in the institutions and habits that tell them apart. In this anniversary year, defending the authentic record of our past is a way of honouring those who risked everything to create it.
Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net




Comments