Deepfake Threats in Africa: How Disinformation Is Going Multimedia

For most of the past decade, disinformation in Africa meant a forwarded WhatsApp message, a doctored screenshot or a recycled photo attached to a false caption. Those formats still circulate. But the false material reaching phones in Lagos, Nairobi and Johannesburg increasingly arrives as video, audio and fabricated broadcast footage, and the gap between fake and genuine has narrowed enough that even trained verifiers struggle.

The shift matters beyond politics. The same tools that clone a presidential voice can also pass a bank’s identity check, which means deepfake threats in Africa now sit across elections, finance and public trust at once.

From text rumours to synthetic video

Researchers describe synthetic media less as a new kind of lie than as a cheaper way to produce and spread one. A July 2026 policy brief from the Policy Center for the New South argues that deepfakes act as a multiplier of existing political and social tensions. It also notes that “deepfake as a service” offerings let political actors with no technical skill commission an operation for a few tens of thousands of dollars.

The continent’s circumstances make that cost advantage count for more. Mobile-first audiences share video and voice notes freely through closed messaging groups, where fact-checkers cannot easily see what is spreading. Forus, a civil society network, documented a June 2025 fake broadcast-style news clip in Nigeria claiming soldiers were guarding cattle in Benue State’s Yelwata community. It also cited research suggesting about 60 percent of disinformation campaigns targeting Africa are foreign-sponsored, with West Africa bearing the heaviest load.

Nigeria’s 2027 election is the test

Nigeria’s presidential election is scheduled for January 16, 2027, and TechCabal reports that experts doubt the country’s institutions are ready for the volume of AI-generated content. A video that circulated on X this year, apparently showing President Bola Tinubu answering a war threat from Donald Trump, was confirmed as a deepfake by Dubawa after fact-checkers noted unnatural body movements. The Independent National Electoral Commission has set up an internal unit to study AI, but the same report says it has not yet rolled out any programmes.

The problem is already visible in smaller contests. During the June 2026 Ekiti governorship election, CDD Fact Check identified an AI-generated image falsely presented as showing voters waiting to sell their votes.

Fact-checkers say the main change since 2023 is not the existence of fake content but its speed, quality and accessibility. Mayowa Tijani of TheCable told Techpoint Africa that audio deepfakes are now cheaper and harder to detect. Audio is also the harder format to check, he said, because few people are building tools to detect it and the work leans heavily on human reviewers.

The liar’s dividend

A less obvious risk is that fakes make genuine evidence easier to deny. At a November 2025 panel on media literacy and the 2027 election, reported by Roundcheck, a speaker described the “liar’s dividend”: a politician can dismiss an authentic recording by claiming it was generated by AI.

This turns deepfakes into a tool for evasion as well as deception. A single convincing fake can cast doubt on every later piece of real footage, and in countries where trust in institutions is already thin, that doubt does not need a sophisticated forgery to take hold.

Fraud is moving in the same direction

The multimedia shift is not confined to politics. Identity verification firm Sumsub’s Identity Fraud Report 2025-2026 found overall fraud rates falling in Nigeria (down 54 percent year on year), Kenya (down 42 percent) and South Africa (down 31 percent), helped by stronger verification. What remains is more sophisticated. Deepfake-enabled attempts to bypass liveness checks are increasingly common, and in Kenya deepfakes account for nearly 10 percent of all fraud attempts.

Sumsub’s survey data, summarised by IT-Online, found that nearly one in five people across Africa had been directly targeted by a deepfake attack, while 24 percent admitted they could not reliably tell fakes from real content. According to a Smile ID report as summarised by a South African security firm, South Africa accounts for 22 percent of the continent’s deepfake fraud cases, the highest share in the study. Those figures come from vendors that sell verification services, so they are best read as indicators of direction rather than precise measurements.

What a response could look like

The policy debate is still early. Commentators in Nigeria have proposed criminalising deceptive AI-generated electoral content, requiring labels on AI-generated campaign material and having INEC work with fact-checkers such as Dubawa, Africa Check and FactCheckHub. One African Liberty commentary points to Ghana’s practice of partnering with platforms to monitor election-period disinformation as a model. These are proposals, not enacted measures, and Nigeria’s broader AI governance still depends on draft bills rather than binding AI-specific law.

Fact-checkers are adapting in the meantime. Some are turning to prebunking, which means giving voters verified facts before a falsehood spreads, and to AI-assisted claim matching to shorten response times, as TechCabal reported in 2025. Even so, the volume of synthetic content is likely to outpace the people verifying it.

Where this leaves African audiences

Deepfake threats in Africa are not a single problem with a single fix. In elections, they erode shared facts; in banking, they exploit automated onboarding; and in everyday conversation, they make audiences less certain about what they see and hear.

The practical gap is capacity. Detection tools tuned to African voices, faces and languages are scarce, electoral bodies are only starting to build expertise, and media literacy remains uneven. Over the next few months, Nigeria’s campaign season will show how much of that gap can be closed before the votes are cast.