South Africa’s 2026 municipal elections are shaping up to be fought not only at ward meetings, on street corners and in community halls, but across the full chaos of social media, where deepfakes, AI-generated clips and political propaganda can spread faster than any press statement can catch up. The problem, as we head toward 4 November 2026, is that the country’s election rulebook is still mostly focused on broadcasters — not the platforms where millions of South Africans now consume politics, gossip and outrage.
That gap matters. Icasa has recently published its updated Municipal Party Elections Broadcasts and Political Advertisements Amendment Regulations 2026, giving the impression of a tighter framework for election season. But in practical terms, the rules mainly bind the SABC, eMedia, MultiChoice and radio stations. They do not directly cover X, TikTok, YouTube, Facebook or Instagram, which is exactly where election interference is now most likely to surface.
The regulator’s move may help tidy up old-school election advertising, but it barely touches the modern threat. And that threat is no longer just crude misinformation. It is increasingly about hyper-targeted manipulation, AI-generated voices, synthetic video and content engineered to inflame local rivalries in a specific ward, not just a whole province. In other words, the battle has moved online, while the rules have stayed in the analogue era.
The Electoral Commission of South Africa (IEC) clearly understands this. Chairperson Mosotho Moepya has been warning since February about a “flurry of deepfakes” and about disinformation shifting from broad national narratives to more granular, ward-level lies. Chief electoral officer Sy Mamabolo has also confirmed that the commission is building internal capacity to respond to social media threats, a sign that election administrators know the risk is already here.
But the warning lights should be blinking even brighter after what has happened elsewhere. In Slovakia in September 2023, a fake audio recording surfaced just two days before polling, apparently showing liberal leader Michal Šimečka discussing election rigging with a journalist. It was a deepfake, released during a 48-hour pre-election silence period, when parties and media had little room to answer. By the time fact-checkers dismantled it, the damage had already been done and Robert Fico’s pro-Russian Smer party had won.
That episode matters because timing is now part of the weapon. Even when a falsehood is exposed, it can still land with enough force to tilt undecided voters or harden the beliefs of those already leaning one way. In the age of algorithmic distribution, a lie does not need to be believed by everyone. It only needs to be seen by the right people at the right moment.
Why the focus keyword matters in South Africa’s election misinformation fight
The South Africa election deepfakes debate becomes even more urgent when you look at the way social platforms are now being used in political conflict globally. In Romania, a once-obscure pro-Russian figure, Călin Georgescu, surged from around 5% in the polls to 23% on election day in November 2024, driven largely by an unexplained flood of TikTok content. The country’s constitutional court later annulled the result on 6 December, the first time an EU member state had thrown out a national election because of social media manipulation.
That is not a distant, abstract warning. It shows how quickly a platform can manufacture political momentum, especially when content is designed to look authentic, emotional and local. The visuals may be fake, but the reaction they trigger is very real. That is what makes the current wave of AI-assisted political content so dangerous: it can make fringe ideas look mainstream in a matter of hours.
The same pattern is visible in the information war around the US-Israel-Iran conflict. A small Iranian group called Explosive Media has been producing AI-generated Lego-style animations aimed at Donald Trump and Benjamin Netanyahu. The clips mock both leaders and have been viewed millions of times across X, TikTok and Instagram. YouTube eventually banned them in mid-April, but the other major platforms did not move in the same way.
Media analyst Marc Owen Jones, who studies propaganda at Northwestern University in Qatar, has described the material as “troll propaganda”. His point is simple: if a country cannot win militarily, it may try to win the meme war. And in a crowded social feed, visual humour often travels further than sober rebuttal.
That is why this latest form of political content is so effective. The Lego-style clips are not deepfakes in the strict sense. They are something different: AI-styled iconography, built for a generation that now consumes politics as content, not just ideology. The message is packaged in a format people recognise instantly, then pushed through platforms that reward outrage, novelty and repetition.
South Africa has already had its own taste of how digital manipulation can distort public debate. Back in 2016, the British PR outfit Bell Pottinger, working for the Gupta family, ran a sustained campaign on Twitter that helped popularise the “white monopoly capital” narrative as a counter-attack against state capture reporting. Investigations later exposed the role of dozens of bot accounts, while journalists such as Ferial Haffajee, Adriaan Basson and Peter Bruce found themselves targeted in coordinated online attacks.
Then came a more brazen sign of what the AI age would look like. In March 2024, Duduzile Zuma-Sambudla circulated a clearly fake, synthesised clip of Donald Trump’s voice endorsing the MK Party. The clip was poorly made, but it still travelled widely enough to matter. That is the key lesson: bad fakes do not have to be convincing to be effective.
The reason they work is psychological as much as technical. Content that confirms what people already believe can strengthen group identity, deepen suspicion of opponents and make supporters more committed. In a political climate as fractured as South Africa’s, that is fertile ground. Service delivery has collapsed in many metros, trust in institutions is low, and parties are increasingly fighting over identity as much as policy.
At the same time, platforms can still be used for genuine political accountability. Helen Zille’s digital campaign for Johannesburg mayor is a good example. Much of it plays out on TikTok and Reels, with the former DA leader appearing in flooded streets, inspecting broken infrastructure and highlighting municipal failure in a way that older media formats simply cannot match. The dysfunction she documents is real, and the reach she gets is far beyond what a traditional party broadcast could deliver.
But that upside comes with a cost. The same tools that can expose potholes and broken traffic lights can also be used to manufacture a false opponent, fake a speech or invent a scandal. By November, any political operator with basic tools and minimal skill could generate a convincing audio clip or video at very little cost. In that sense, Icasa’s election rules are aimed at the wrong battlefield.
Our view is that the real fight now sits with the platforms themselves, especially those that have resisted meaningful election transparency measures. A 2024 cooperation agreement with Meta, TikTok and Google was a step forward, but X, under Elon Musk, did not join in. Without stronger disclosure requirements for synthetic political material, and without an enforceable arrangement with the major social platforms, the IEC risks spending the campaign season chasing one fake after another while the next one is already trending.
South Africa’s next municipal election will not just test parties and candidates. It will test whether our institutions can keep up with the speed of online manipulation. The country needs faster detection, clearer labelling of synthetic content and real platform accountability before the campaigning gets fully underway. Otherwise, by the time a lie is corrected, the damage will already be part of the election outcome.