SA has seen a sharp increase in AI deepfake fraud, according to LexisNexis Risk Solutions, which warns that the trend could undermine efforts to rebuild investor and institutional trust following the country’s recent exit from the Financial Action Task Force (FATF) grey list.
The fraud prevention company has recorded a 180% year-on-year increase in attacks globally and says deepfakes are now one of the fastest-growing fraud vectors worldwide. It cites Juniper Research, which predicts that 100 billion identity-related checks will be carried out this year and that one in every 100 failed checks will contain a deepfake.
Kimberly Sutherland, global head of fraud and identity at LexisNexis Risk Solutions, says deepfakes vastly complicate digital identity verification. "Protecting against this surge of attacks requires a solid line of defence incorporating end-to-end capture, fraud analysis and liveness checks. Even the smallest gap in your defences is like an open window that a fraudster can climb through.”
The company warns that AI-generated deepfake documents, images and liveness videos could leave organisations exposed if their identity and customer onboarding checks fail to keep pace. It expects businesses to face increasing volumes of daily attacks targeting their digital services, particularly in a market like SA.
Sutherland says SA is a classic example of a fast-growing digital economy driven by a young, dynamic population and an advancing payment infrastructure. "This makes the environment ripe for attack: younger generations tend to be less secure when interacting digitally; organisations are rushing to adapt and provide services quickly; and payment environments allow the fast movement of money following an attack, all of which benefits fraud networks.”
She adds that fraud is inherently linked to money laundering – fraud cannot function without a network of mules to launder the proceeds. “Compliance with FATF and other regulators requires a detailed understanding of these links, yet often the focus is on compliance, and the fraud side can be overlooked.”
According to LexisNexis Risk Solutions, bad actors use deepfakes to bypass identity checks and create new accounts or take control of existing ones to make unauthorised payments, withdrawals and online purchases, launder the proceeds of crime or abuse new customer bonus incentives.
The company’s latest cyber crime report found that one in every 11 new account creations in 2025 was a fraud attack, and that almost a fifth of all reported fraud involved unauthorised access to customer accounts.
As deepfakes become more realistic, identity checks need to be capable of spotting nuanced flaws in document security features and closely examining facial expressions and skin tone.
“Highly realistic deepfakes call for forensic examination of hundreds of security features: document structure, image integrity, holograms, etching and microtext," says Sutherland. "Deepfakes typically fail on several minor flaws, as opposed to physical forgeries that fail on one major issue, but they are not easy to spot with the human eye during manual checks. The same goes for deepfake images and videos. Checks need to assess micro movements in facial muscles, analyse light reflection and detect image manipulation and injection tactics.”
Shane O’Sullivan, research analyst at Juniper Research, says digital identity verification is evolving, with the core requirements shifting towards the technical ability to integrate multiple trust signals into a coherent system architecture. "Effective solutions depend on the co-ordination of document authentication, biometric liveness detection and real-time risk analysis within a single workflow. Increasingly, fraud detection system success is defined by how well it can detect advanced threats such as synthetic identities and deepfakes while maintaining interoperability across standards and minimising latency and user friction.”
Sutherland says the risk to businesses is real from both a financial and reputational standpoint. "The reality is that AI-generated attacks are practically doubling year over year and getting more sophisticated with every attack.
“It’s true that organisations are being caught out by this sharp increase in the sophistication of fraud attacks," she says. "Unfortunately, there is no shortcut to strong fraud prevention. Organisations need to adopt the latest controls relevant to their customer journey, including – where necessary – those that utilise AI and advanced techniques to spot deepfakes and manipulation within their ID verification processes. The good news is that these solutions are not always costly and, if implemented in the best way, can significantly help reduce fraud losses and improve operational efficiency, ultimately helping reduce fraud prevention costs.”
Share
