Detect Deepfakesby Resemble AI
Deepfake case study · Multi-modal

Real-Time AI Deepfakes Enable Live Interactive Impersonation of Public Figures on Streaming…

Real-time AI deepfakes now enable sophisticated live impersonations on streaming platforms and video calls that risk financial loss and disinformation

Incident date
Jan 2024
Target
CFO of a Hong Kong finance company
Updated Jun 13, 2026 · 2 min read

Real-time generative models now allow attackers to overlay a target's face and voice onto live video feeds with latency low enough to support interactive, conversational responses. These synthetic impersonations enable bad actors to influence viewer perceptions or solicit fraudulent payments while appearing as a trusted authority.

What happened

Modern deepfake technology has evolved from static image manipulation to live, interactive impersonation. Operators utilize public recordings to train models that drive synthetic figures in real time, matching lip synchronization and facial expressions to spoken output. This allows the impersonator to reply directly to incoming comments or questions, making the interaction feel authentic to the viewer.

On streaming platforms, these synthetic identities exploit existing features like virtual gifts and paid messages, which direct financial resources to the impersonator rather than the actual individual. Because these interactions occur in real time, statements made by the AI receive immediate attribution to the target, often before independent verification can take place. Detection remains difficult as visible artifacts such as hair movement or background consistency vary based on lighting and operator skill.

In corporate settings, this technology has facilitated high-stakes fraud. In 2024, a finance employee in Hong Kong participated in a video conference with a deepfaked CFO and other colleagues, resulting in a loss of approximately $25 million. A similar incident occurred in 2025 involving a Singaporean finance director, who was defrauded of $499,000 via a Zoom call and WhatsApp, though authorities were able to recover most of those funds. These attacks rely on a combination of reconnaissance, phishing, and the creation of artificial urgency to pressure employees into making rapid payments. As these tools become more accessible and refined, they pose a significant threat to organizational security, prompting new regulatory scrutiny, such as the UK’s Economic Crime and Corporate Transparency Act, which holds large firms accountable for failing to prevent such fraud.

Sources