South Africa Has Africa's Highest Deepfake Fraud Rate. Your Board's Marketing Partners Are Next.
The Guardian has just reported deepfakes destroying influencer credibility globally. South African data shows the same mechanism is already here, and it is accelerating.
Sign up for my Substack daily AI newsletter here.
See my AI Training course portfolio for corporate Business Leaders here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
The Guardian published a story recently that most South African boardrooms will read as an American problem. A California lifestyle influencer named Emily Schuman discovered an AI-generated version of herself in an advert for a telehealth company she has never worked with. Molly Tranchin, a fashion influencer, described a running fight against deepfake versions of her likeness selling products she has not endorsed. Taylor Swift, Kim Kardashian, Oprah Winfrey, Tom Hanks and Gwyneth Paltrow have all been targeted. The Surfshark research team estimates that global consumers may have lost as much as $3.7 billion to deepfake scams in 2026, with social media impersonations accounting for about half of all losses.
That is the American reading of the story. The South African reading is different. The specific mechanism the Guardian describes has been operating here for at least two years. Our fraud data show it is doing more damage in South Africa than anywhere else on the continent. And the next targets in the sequence are the corporate ambassadors, executives and named human faces that your organisation puts on its own marketing.
CONTEXT AND BACKGROUND
The Guardian story sits inside a well-documented global trend, but the South African numbers are worse than the framing suggests. The 2026 Digital Identity Fraud Report by identity verification company Smile ID found that 22 per cent of fraud cases in South Africa now involve AI-generated impersonation, the highest share on the continent. Almost nine in ten rejected verification attempts in the region are now linked to AI-assisted impersonation and spoofing during biometric checks. Some 47 per cent of cases involve no-face-match impersonation, and another 40 per cent are spoofing attacks designed to defeat liveness detection and facial recognition systems.
Specific cases in this country now form a pattern. In 2024, Leanne Manas, the SABC broadcast anchor, was deepfaked to endorse weight loss products and online trading platforms on Facebook and TikTok. A South African doctor’s likeness was used in deepfakes to promote non-existent male-potency products. Banxso, a broker licensed by the Financial Sector Conduct Authority, saw its brand hijacked in deepfake advertisements on South African social media. More recently, the Financial Sector Conduct Authority has warned about deepfake videos impersonating President Cyril Ramaphosa and Patrice Motsepe to promote fake investment opportunities. The Elon Musk deepfake targeting South African investors was successful enough that a significant number of South Africans transferred money to fraudulent schemes on the strength of a fabricated endorsement. This is not a future risk. It is a live commercial and reputational risk with named victims and quantified losses.
A broader signal in the same direction is coming from the marketing industry itself. AI-generated virtual influencers accounted for $1.37 billion of global brand spend in 2026, according to the marketing consultancy Digital Applied, or roughly 4 per cent of the total influencer marketing budget. That is not fraud. It is a legitimate marketing channel. But it is a channel whose economics are built on the same underlying technology that is being weaponised against the influencers the Guardian article names. The market is signalling that synthetic humans are commercially viable content. The fraud economy is signalling that the same synthetic humans, without authorisation, are commercially viable weapons.
INSIGHT AND ANALYSIS
The pattern that has just reached global influencers is the same pattern that reached South African financial services 24 months ago. The mechanism runs in the same order. Somebody with a public face becomes the target because the material required to model them is already on the internet, in interviews, photographs, television appearances and social media. A synthetic version of that face is then attached to a proposition — an investment scheme, a weight-loss product, a fake advertisement — and pushed to an audience through social media platforms that have neither the incentive nor the technical capacity to catch it before it reaches the reader. The proposition converts because the face is familiar and the platform where it appears is one the target audience already trusts. By the time the deepfake is reported, taken down, or publicly denied, the money is gone, and the damage to the reputation of the person whose face was used is already done.
The deepfake economy is one specific version of what those companies are worried about. The generative tools that allow this mechanism to run at scale are being built and released faster than any of the platforms hosting the resulting content can identify and remove it, and faster than any of the legal systems adjudicating the harm can meaningfully respond. South Africa is not the exception in this. It is the leading edge, at least on the continent, and by a specific data point of 22 per cent it is doing more badly than any of its African peers.
The consequence for a South African board is specific. If the mechanism has already worked against the sitting President, against one of the country’s most recognisable business figures, against a SABC anchor and against an FSCA-licensed broker, it will work against your organisation’s brand ambassadors, your executives who appear on television, your senior partners whose names carry weight with customers, and your marketing collateral that features named human faces. The question is not whether. The question is when, and how prepared the organisation is when it happens. Most South African corporate risk registers do not currently carry deepfake and generative AI content risk as a named item. Most audit committees have not received an update on it. Most marketing budgets are being deployed without a specific contractual clause on synthetic-content provenance. That combination is now materially incomplete, and the Guardian story is the international signal that ought to prompt the change.
The international response framework has already taken shape. The Coalition for Content Provenance and Authenticity, known as C2PA, is a standards body founded in 2021 by Adobe, Microsoft, the BBC, Intel and others. Its Content Credentials specification embeds cryptographically signed metadata directly into a media file, recording who created it, when, what tools were used, whether AI was involved, and every meaningful edit since capture. The specification is now supported by OpenAI, Google, Meta, Amazon, Sony, Nikon, Canon and Samsung. The US National Security Agency issued formal guidance recommending its use in January 2025. Since 2 August 2026, EU AI Act Article 50 has required machine-readable content marking for AI-generated content, and C2PA is the technical standard the industry has coalesced around to meet that requirement. Boards in Europe already have a defensible answer to the question of what their organisation is doing about synthetic content risk. Boards in South Africa do not yet have one, because the national conversation has not yet begun.
IMPLICATIONS
For South African corporate boards, the immediate implication is that the risk register needs an addition and the audit committee agenda needs a standing item. The specific questions the board should be asking are three. First, which of the human faces our organisation uses in marketing, executive communications and customer engagement are exposed to this risk, and what is our current visibility into unauthorised uses of those faces? Second, what contractual protections do we have with our marketing agencies, our influencer partners, and our media buying platforms that would give us recourse if a synthetic version of our brand or one of our people appears in an advertisement we did not commission. Third, what is our incident response capability if a deepfake using our organisation, a partner, or a key executive appears tomorrow, and who is accountable for that response?
For South African policymakers and regulators, the implication is that the Financial Sector Conduct Authority and the Prudential Authority need to move from case-by-case warnings to a systemic response. The FSCA’s warnings about specific Ramaphosa and Motsepe deepfakes are appropriate but reactive. What is now needed is the equivalent of the cybersecurity regulatory posture the sector has developed over the past decade, applied to synthetic content and identity impersonation risk. That is not simple, and it will require legislation. There is currently no recognition of image rights in South African case law or legislation, which places our citizens and our companies in a materially weaker legal position than their counterparts in the European Union, California, or China. That gap is now costing money and reputation, and it will cost more.
CLOSING TAKEAWAY
The Guardian story is not an American problem read at a safe distance. It is a description of a mechanism that has already succeeded against the country’s most recognisable public figures, that our own identity verification industry now measures at 22 per cent of all fraud, and that will succeed next against the corporate faces on your organisation’s marketing. The international response framework exists. The domestic conversation has not started. The board pack that walks into the next audit committee meeting with an AI strategy paper and no deepfake risk update is walking in with an incomplete answer. It should begin this quarter.
Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net
Comments