On the morning of November 2, 2026, the Gallagher Convention Centre in Midrand looked like the engine room of South African democracy. It was the national results centre of the Independent Electoral Commission, a place where the huge, messy, hopeful act of voting is turned into something orderly and shareable. Outside, security guards checked the passes of journalists and party agents who streamed through the sleek entrance. Inside, the air was thick with the smell of coffee and the sound of foot traffic, murmuring voices, and the soft hum of screens showing live counts. This is the kind of place where democracy tries to prove that it is still real, still transparent, still trustworthy. But on this particular morning, the people in the room were thinking about something else. Sy Mamabolo, the chief electoral officer, stepped up to the microphone with the composed but careful expression of someone carrying extra weight. He had not come simply to talk about tallies and voter turnout. He had come to deliver a warning. The IEC had received more than 700 complaints related to disinformation in the election period. Among them, Mamabolo said, were reports of AI-manipulated content. The room went quiet. It was one of those moments where the future stops pretending to be hypothetical. The question was no longer whether artificial intelligence could affect elections. It was already here, arriving in the feeds of ordinary South Africans, dressed in the audio and video likeness of real people. The battle for this election was no longer happening only at polling stations and counting centres. It was also happening on phone screens, in social media timelines, and inside the strangely persuasive glow of a deepfake that looked and sounded exactly like someone you trusted.
More than seven hundred complaints. It is the kind of number that can slip past the mind, but it deserves to sit with us for a while. Seven hundred people, or perhaps more, were disturbed enough, confused enough, or angry enough to pause their lives and report something to the electoral commission. Some of these complaints are about old-fashioned lies: false claims about voting stations, invented stories about ballot boxes, rumours about results that had never been announced. But the most unsettling reports, according to Mamabolo, involve content produced or changed using artificial intelligence. We are not talking about clumsy Photoshop edits or obvious fake screenshots. We are talking about videos that look real, voices that sound authentic, images that resemble genuine photographs. A deepfake might show a candidate making a racist comment they never made. A cloned voice might say, in the candidate’s own language, that the voting day has been moved to next week. A manipulated image might depict chaos outside a polling venue, where nothing actually happened. Mamabolo said that some of this content had been circulating on X, the platform formerly known as Twitter. That is important, because X is a public square where things spread faster than truth can chase them. A single post can be shared thousands of times before the first fact-check is even published. And the people sharing it are often not malicious. They are our neighbours, our mothers, our old school friends, people who see something alarming and hit the share button out of a misplaced instinct to protect others. That is the cruelty of AI-manipulated content: it takes our humanity, our desire to connect and warn, and turns it into a delivery system for lies.
Sy Mamabolo is not a celebrity. He is a public servant in the truest, least glamorous sense. Standing at the podium, he looked tired in the way that people look when they have spent months thinking about things that can go wrong. He did not raise his voice. He did not need to. His words were enough. He said that the commission was aware of the problem, that it was taking the complaints seriously, and that it was working to verify and respond to each one. There is something deeply difficult about that job. Mamabolo is now in the strange position of knowing that his own face and voice could be used against him. A citizen could be shown a video of the chief electoral officer announcing that the election had been cancelled, and if the technology is good enough, every tiny movement of the mouth and every inflection of the voice would look right. The only defence is vigilance. At the Gallagher Convention Centre, teams of electoral officials have become digital detectives. They do not just monitor spreadsheets and vote-counting software. They monitor timelines, video clips, audio files, and websites. They ask questions like: does the shadow on that face match the light in that room? Does the blink pattern look human? Does that voice have the same breathing rhythm as the real speaker? This is not the job that they signed up for many years ago, but it is the job now. And Mamabolo made it clear that the IEC would not shy away from it. He thanked the people who filed complaints. He encouraged more people to do the same. That might sound bureaucratic, but it is not. It is an invitation to share the burden of protecting the truth.
This is not a uniquely South African problem. Around the world, election officials are facing the same unsettling reality. Artificial intelligence has made it cheap and easy to create convincing false content. You do not need an editing studio or advanced skills. You need an idea, a target, and a little time. The implications for democracy are enormous. Imagine a voter in a small town, exhausted from work, scrolling through social media late at night. They see a video that appears to show a locked polling station with a sign saying the election has been moved to another city. They do not fact-check. They have never needed to fact-check a video before. They simply stay home on voting day. That one fake video has quietly erased their voice. Imagine another voter receiving a voice note from a relative, urging them not to vote because the results are already rigged. The voice sounds familiar, but it is not the relative; it is a synthesis of their voice created by software. The feeling of betrayal and hopelessness is real, even if the message is fake. This is what AI-manipulated content does. It does not need to convince everyone, forever. It only needs to create doubt in the right moment, in the right minds. A false rumour can linger for days. A fake video can be watched and shared and debated until the nuance of its falsehood is lost in the noise. Around the world, elections are being changed by this shadow war. And in Midrand, in the calm, brightly lit halls of the results centre, the war had reached South Africa.
But there is a counterbalance, and it starts with that number: seven hundred complaints. The fact that those reports exist is not just a warning; it is a sign that the system is alive. Mamabolo and the IEC have built a mechanism for citizens to raise their hands and say, “I saw something, and it does not feel right.” That mechanism matters more than we might think. Once a complaint is registered, it can be investigated. Patterns can be identified. A platform can be contacted. A video can be traced back to its source. A lie can be confronted. Mamabolo said the commission was engaging with X and other social media platforms, even though those conversations are often slow and frustrating. Meanwhile, ordinary people have more power than they realise. We do not have to be experts in deepfakes to protect ourselves. We simply need to slow down. Before sharing something shocking, we can ask: who made this? What is the source? Does it appear anywhere official? Maybe we can run a reverse image search. Maybe we can wait ten minutes before sharing. That small pause, that habit of hesitation, is the thing that manipulators fear most. They want us to move on impulse, to feel outrage or anxiety before we have time to think. The human firewall, made up of sceptical, curious citizens, is still the best defence. And when something truly seems wrong, we can do exactly what those 700 complainants did. We can report it. We can tell an official channel. We can warn our friends with a source, not just a screenshot. That is what it means to be part of a democratic community in the age of artificial intelligence. We cannot stop every lie from being created. But we can make sure that lies do not travel under the cover of our silence.
In the end, the true story of this election will not be told only by the numbers on the screens at the Gallagher Convention Centre. It will be told by the choices made by millions of South Africans as they pick up their phones and look at the world. The more than 700 disinformation complaints are not proof that the system is broken. They are proof that thousands of people are paying attention, that they care enough to look twice, and that they are willing to do something. Mamabolo’s briefing was a reminder that free and fair elections are not automatic. They are built one piece at a time, by officials who refuse to rush, by citizens who refuse to be fooled, and by ordinary conversations around dinner tables, in taxis, at workplaces, where someone says, “Wait, have you actually checked this?” The deepfakes will get better. The technology will become more convincing. But human beings are not as passive as the algorithms assume. We have memory, context, relationships, and common sense. We have the ability to ask questions and the courage to admit when we do not know. That is not a weakness. That is the oldest, most powerful tool we have. As the lights stay on at the results centre and the final votes are counted, the fight for truth will continue beyond the headlines. It will continue in every feed, every group chat, every moment of hesitation before the share button is pressed. Democracy was never meant to be easy. It was meant to be watched over. And in this strange, noisy, artificial century, the watchful are the ones who will keep it safe.