Fraud techniques are becoming more advanced, especially with the use of synthetic identities that are increasingly realistic. According to a 2025 analysis, fraud attempts using deepfakes—AI-generated videos or images that mimic real people—had surged by 700 percent in one year, while AI-generated synthetic identity documents had increased by 281 percent. These synthetic identities are often created using real data, such as stolen personal information, making them harder to detect. Most identity verification systems currently rely on a "linear trust model," where an identity document is checked against a live video or selfie. The system then compares the face and document geometrically. Additional "liveness proof" checks, such as blinking or head movement, are used to confirm a real person is present. However, these methods are increasingly vulnerable to sophisticated attacks. Real-time deepfakes and video injection techniques are now being used to bypass these systems by inserting fake video streams directly into the verification process. These streams, often generated through virtual cameras or software layers, appear legitimate to the system before any biometric analysis even begins. Current liveness proof mechanisms are limited by their reliance on simple signals like blinking or head movement, which can now be simulated with high realism. Another major issue is that many systems only analyze the visible image without checking if the video stream itself has been tampered with earlier in the process. To address this, some advanced security systems are adding video injection detection mechanisms. These analyze signs of virtual cameras or anomalies in how images are rendered and synchronized. As fraud becomes more complex, traditional identity verification systems that rely on independent checks—such as document analysis, facial recognition, and device checks—are no longer sufficient. Deepfakes and synthetic data can now only be detected by correlating multiple layers of analysis, including biometric consistency, device fingerprints, network data, and video behavior. This shift is changing the role of fraud prevention systems, which now need to determine whether a video stream was created authentically rather than just detecting visible flaws. The identity document, once a cornerstone of digital verification, is also reaching its limits. Fraudsters often use real or partially real documents, and in some cases, individuals are coached or used to pass identity checks, similar to financial mules. These developments highlight the need for more robust solutions. The European AMLR regulation, which took effect in July 2027, requires enhanced verification methods beyond video checks, including the use of electronic identities and qualified trust services. While video verification will remain important in high-risk situations, it is evolving into one part of a broader, interconnected security system. This shift marks a new era in identity verification, requiring companies to adapt to a more complex and layered approach to security.