So what are deepfakes and where did they come from? The term "deepfake" emerged in 2017 when a Reddit user published explicit face-swapped videos of celebrities. Since then, deepfake generation tools have proliferated dramatically. Armed with nothing more than a smartphone and internet access, individuals can now produce sophisticated deepfakes during live video interviews or meetings, using readily available, inexpensive or even free software.

Recent advances in artificial intelligence, particularly Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and Diffusion Models, have made it remarkably simple to generate convincing face-swap videos or deepfakes. Technologies once confined to research laboratories are now accessible tools for sophisticated interview fraud.

These deepfake videos seamlessly replace one person's identity with another, showing individuals saying or doing things that never occurred in reality, and creating opportunities for fraudulent activities such as candidate impersonation and other deception schemes targeting recruitment processes.

While deepfakes are often used for legitimate creative and entertainment purposes such as meme videos, humorous clips of celebrities or politicians performing viral dances, or digital resurrection of historical figures in documentaries, their rapid proliferation blurs the boundary between reality and fabrication, posing ethical, social, and security risks.

Despite growing awareness, legislation hasn't kept pace with the evolution of deepfakes, creating a vacuum that forward-thinking organisations are now filling. For example, NextGen Interviews is at present the first and only one-way interview platform to integrate deepfake detection, while competitions like The Global Multimedia Deepfake Detection Challenge tackle broader detection challenges at scale.

Academic researchers have responded with equal urgency. Google Scholar data (illustrated below) reveals a sharp acceleration in published research on deepfake face-swap detection, reflecting the technology's broader societal importance and demonstrating why recruiters and hiring teams are increasingly prioritising this safeguard as best practice.

Research and advisory firm Gartner projects that by 2028, one in four job candidates worldwide will be fraudulent. This projection is further evidenced by a Resume Genius survey of 1,000 hiring managers across the United States, which found that 17% have already encountered candidates using deepfake technology during video interviews.

"As hiring teams realise that creating deepfakes for video interviews requires nothing more than a static image or brief video clip of another person, the question shifts from 'do we need deepfake detection?' to 'why aren't we using it already?' said Luke Cruz, founder of NextGen Interviews and an early proponent of AI integration in business workflows.

"Video interviews have become standard practice globally. They're efficient, convenient, and perfectly suited to how we work today. But conducting them without deepfake detection is like driving without a seatbelt. It's not worth the risk," Luke said.

As deepfakes advance in both sophistication and availability, the threat constantly evolves. Whilst no deepfake detection model is 100% foolproof, the statistics are unambiguous, the technology is accessible, and the threat is real, making the implementation of deepfake detection protocols an urgent operational imperative for organisations of all sizes.

For recruitment professionals, HR managers, and business owners conducting video interviews, implementing deepfake detection has become a fundamental component of due diligence that can no longer be overlooked.

Learn more and start your free trial at nextgeninterviews.ai.