IEC Flags AI-Generated Mandela Video Ahead of Local Elections

An AI-generated video of former president Nelson Mandela urging voters to turn away from the ANC circulated less than a week before South Africa's 2026 local government elections.^[2]

South Africa's Electoral Commission (IEC) has reported it to Real411 and warned voters to be cautious with AI-generated political content.^[1][2]

What Happened

In the video, a likeness of Mandela says: "stop voting for the ANC; they have failed you. They are corrupt."^[1][3] The video also claims the ANC stole pension funds.^[3]

These claims come from a fabricated video. Mandela did not make these statements.

Real411 receives and assesses complaints about suspected election-related disinformation.^[2] Members of the public and candidates can report suspected disinformation there within 36 hours of discovery.^[1]

The video spread widely on social media. Some viewers read it as satire. Others were troubled by how convincing it looked.^[3]

A 94-year-old woman, shown the video by her grandchild, appeared to believe it was real. She said she would not be voting for the ANC.^[1][3]

IEC spokesperson Kate Bapela said the commission is aware of public concerns about the video.^[2] "The commission does not seek to regulate legitimate political opinion or determine voters' political preferences," she said.

Its responsibility, she said, is to safeguard the integrity of the electoral process.^[2]

The IEC said political discourse should fall within the scope of its Disinformation Code, introduced on 4 September 2026.^[1][3] Section 11 requires parties and candidates to monitor their online platforms and "take rapid action to publicly correct and limit the further spread of the disinformation."^[1]

The Code also requires parties and candidates to clearly label AI-generated content. It protects satire, parody and political commentary that fall within protected speech.^[2]

The IEC said it will consider the matter through the relevant processes and legal framework.^[2]

Why It Matters

The 94-year-old viewer shows the risk in plain terms. A familiar face. A fabricated message. A stated voting intention that followed.

The IEC's position is narrow. Satire, parody and criticism are protected. Content that could mislead voters is where it draws the line.^[3]

That makes labelling central. A labelled parody reads as commentary. An unlabelled one can travel as fact once it leaves the account that posted it.

The pressure is wider than one clip. The IEC said it has received more than 700 complaints of election-related misinformation, most linked to posts on X.^[2]

Earlier in October, it teamed up with Google, Meta, Microsoft, TikTok and Moxii Africa to combat disinformation.^[2]

What to Do

• Check who posted it. A party, a candidate, a news outlet and an anonymous account carry different weight. Trace a clip to its first source before trusting it.
• Look for a label. Under the Code, parties and candidates must clearly label AI-generated content.^[2] A political video with no label is a reason to slow down.
• Verify before sharing. The IEC urged the public to check information before passing it on and to consult official Commission channels.^[2]
• Report it. Suspected election disinformation can be reported to Real411 within 36 hours of discovery.^[1]
• Talk to family. Older relatives may not expect a familiar face to be synthetic. Show them that convincing is not the same as authentic.

Verify What You See and Hear

Upload any suspicious image or video frame at witness.vision/scan to check for signs of AI generation. Witness flags content that may be AI-generated. Treat the result as a signal, not proof.

The Witness Chrome extension scans images directly on any webpage.

Scan now at witness.vision/scan