Banks Face Surge in Deepfake Fraud Across the Middle East

Financial institutions across the Middle East are grappling with a sharp increase in fraud driven by synthetic identities and AI-powered deepfakes, placing traditional verification systems under intense strain.

Rob Woods, director of fraud and identity at LexisNexis Risk Solutions, says fraudsters are now exploiting a mix of demographic complexity and regulatory fragmentation in the region to bypass security checks. The region’s high number of expatriates and frequent cross-border mobility provide both cover and confusion in identity verification. Deepfake tools—once primitive—are now capable of creating near-perfect overlays for video and audio, making manual inspection of documents largely ineffective.

AI-based detection systems are increasingly being adopted by major banks across the Gulf, incorporating behavioural biometrics, device fingerprinting, and neural network-driven image analysis. These tools analyse how a user interacts with a device, such as typing patterns or swiping behaviour, to flag irregularities. Woods argues that combining such systems with deepfake detection offers stronger defences than relying on static identity checks alone.

UAE has emerged as a leader in collaborative fraud intelligence, with banks forming consortiums to share risk data and alert each other to emerging threats. Although sharing of fraud expertise remains limited elsewhere due to privacy laws and competition among financial institutions, regulatory bodies are under pressure to harmonise laws and encourage cooperation.

Emerging fraud typologies include authorised push payment scams, impersonation schemes and misuse of synthetic identities to open accounts or conduct high-value transactions infrequently enough to avoid triggering standard risk profiles. Fraud rings are operating more like organised entities, with layered operations spread across jurisdictions.

Academic research underpins many of the new detection methods being trialled. One recent paper describes a hybrid deep-learning model combining recurrent neural networks, transformer encoders and autoencoders to detect transactions that deviate from typical behaviour, achieving high accuracy and precision. Another study introduces audio watermarking frameworks to authenticate voice content and counter deepfake impersonations, a particularly acute problem for call-based customer service fraud.

Regulators in the Middle East and international organisations alike are calling for stronger standards to govern AI tools and verification systems. Transparency, explainability, and consistent audit requirements are emerging as priorities. Some proposals suggest mandatory labelling or traceability of synthetic content, stronger identity verification protocols and shared regulatory standards between countries.