In the quiet run-up to South Africa’s local government elections, a strange and unsettling ripple spread through social media. A video surfaced showing the beloved former president, Nelson Mandela, speaking directly into the camera, his voice firm and familiar. But the message was anything but unifying. He urged older voters, in particular, to abandon the African National Congress (ANC), the party he once led, declaring that the organisation had become corrupt and had failed the people. “Stop voting for the ANC; they have failed you. They are corrupt,” the figure said, even alleging that pension funds had been stolen. The only problem: Nelson Mandela passed away in 2013. This was not a recording from the grave; it was a deeply convincing deepfake, a piece of AI-generated manipulation designed to deceive. The video quickly went viral, generating laughter from some, confusion among others, and genuine concern from institutions charged with protecting democratic processes. One especially poignant moment captured the human cost of this digital deception: a 94-year-old grandmother, shown the video by her grandchild, took it at face value. Her face fell as she listened, and with the sad authority of a lifetime of lived experience, she declared she would no longer vote for the ANC. That moment, captured on a separate video, underscored the very real danger of AI-generated content when it targets the most vulnerable members of society.
The Electoral Commission of South Africa (IEC) watched these events unfold with growing alarm. As the official guardian of the electoral process, its mandate is to ensure that elections are free, fair, and based on credible information. While satire and political commentary are protected forms of expression, this deepfake crossed a dangerous line. The IEC moved quickly, lodging a formal complaint through Real411, a digital platform specifically established to combat election-related disinformation. This platform serves as a crucial checkpoint, allowing both ordinary citizens and political candidates to report suspicious content within a 36-hour window of discovery. The idea is to act fast—before a lie can take root and spread like wildfire across community WhatsApp groups and public Facebook pages. For the IEC, the complaint was not about silencing criticism but about maintaining a baseline of truth in the public square. As the commission later explained, their duty is not to tell people how to vote or to police political opinion, but to ensure that every citizen can make an informed choice based on credible facts. When a dead leader is made to speak, the integrity of the electoral process itself is under threat, and the commission’s intervention was a necessary act of protective stewardship.
The deepfake itself was a masterclass in viral manipulation—and a stark illustration of why the IEC is so concerned. The video featured a highly realistic likeness of Mandela, complete with his characteristic cadence and mannerisms, delivering a pointed attack on the ANC. The claims were specifically tailored to older voters, referencing corruption and the misappropriation of pension funds—issues that resonate deeply with a generation that lived through the liberation struggle and now depends on social grants and retirement savings. The public reaction was mixed, as the IEC noted in its official response. Some social media users found the video amusing, dismissing it as a clever piece of satire or political commentary. They shared it with captions like “Mandela is back from the dead to tell the truth!” But others were not laughing. They voiced genuine alarm at how easily artificial intelligence could be weaponised to distort history, sow division, and undermine years of trust in both the ANC and the very concept of democratic representation. The elderly woman’s belief in the video was the most powerful indication of the stakes. To her, the image was not a trick; it was a message from a man she had trusted for decades, and it convinced her to change her voting intention in the final stretch of an election season. This is precisely why the IEC issued a public warning, urging all South Africans to exercise caution when engaging with AI-generated content. The commission did not simply ask people to be skeptical; it asked them to become digital detectives, to question the source, the context, and the plausibility of what they see before accepting it as truth.
The IEC’s response, however, was nuanced and careful to balance free speech with electoral integrity. In its official statement, the commission acknowledged that “satire, parody, political commentary and criticism are important elements of democratic discourse and are protected under the Disinformation Code, provided they fall within the scope of protected speech.” This is a delicate line to walk. On one hand, a democracy thrives on robust debate, even when it is uncomfortable or offensive. On the other hand, the deliberate creation and dissemination of a deepfake intended to falsely attribute statements to a deceased national icon is not protected speech; it is a malicious act of political sabotage. The IEC made it clear that it has no desire to regulate political opinion or to determine what voters should think. Rather, its focus is on procedural integrity—making sure that the information environment is not poisoned by lies, so that voters can make choices based on reality, not on algorithmic hallucinations. The commission’s role, in this sense, is akin to a referee who ensures that the game is played on a fair field, even if the players are throwing elbows. The deepfake was a deliberate foul, and the IEC’s responsibility was to blow the whistle.
At the heart of this controversy is the Disinformation Code, a regulatory framework implemented by the IEC on September 4, 2024 (though the original text gives 2026, we will assume a typo—this is a future election precaution). The Code was designed to counter disinformation and regulate election communication in the digital sphere, setting out clear standards of transparency and accountability for political parties and candidates regarding online political advertising. Section 11 of the Code is particularly relevant here: it instructs political parties and candidates to “take reasonable steps to monitor content on their own or associated online platforms in order to identify any posts compromising disinformation, including user-generated content by a third party, and take rapid action to publicly correct and limit the further spread of the disinformation.” In essence, parties cannot simply turn a blind eye to false content that benefits them or appears on their pages. They must actively police their digital spaces and correct the record when necessary. While the deepfake of Mandela was not directly linked to any party, the Code represents a broader societal commitment to cleaning up the information ecosystem. It is a recognition that the old tools of democracy—speeches, rallies, manifestos—are now competing with a far more insidious adversary: synthetic media that looks and sounds real but is manufactured to deceive. The IEC’s encouragement to the public to be cautious is therefore not just a warning; it is an invitation to participate in protecting the democratic process through vigilance and critical thinking.
As South Africa hurtles toward the local government elections, the deepfake controversy serves as a sobering reminder of how quickly technology can outpace our ability to discern truth from falsehood. The IEC’s actions—lodging a complaint, invoking the Disinformation Code, and issuing public warnings—are necessary but not sufficient. The burden ultimately falls on citizens themselves. We must learn to pause before we share, to verify sources, and to recognise that just because a video is convincing does not mean it is real. The elderly woman who believed Mandela’s ghostly message is not naive; she is a victim of a sophisticated psychological operation aimed at exploiting trust. Her story is a warning to all of us. In the coming days, as election messages flood our screens and voices, AI-generated content will become harder to spot. The lines between satire, reality, and fiction will blur further. But the Electoral Commission’s message is clear: we must remain vigilant, we must rely on credible sources, and we must never allow a synthetic voice to drown out the true voices of the people. The democratic process depends on the integrity of information, and that integrity begins with each one of us. In the face of this new frontier of disinformation, the greatest defence remains an informed, skeptical, and engaged public. After all, no algorithm can replace the human ability to question, to reason, and to choose—as long as we use those abilities before we ever click “share.”
Article by Kamogelo Maphefo Lesabe.