The rise of deepfakes and mistrust in the digital age

In early 2022, a video circulated showing Ukrainian President Volodymyr Zelenskyy telling his soldiers to lay down their arms and surrender to Russia. Within hours, experts identified it as a deepfake, an artificially generated content designed to fabricate reality. Though quickly debunked, the video briefly caused confusion among Ukrainian citizens that were already under extreme stress.

This was not the only incident related to deep fake. As AI technology advances, deepfakes have evolved from being an innovation to genuine threat and in cases, psychological weapons, undermining our ability to distinguish truth from fiction in an already fragmented digital world.

What exactly are deepfakes?

Deepfakes use deep learning algorithms, a subset of artificial intelligence, to create hyper-realistic fake videos, images, or audio showing real people saying or doing things they never did. The technology behind deepfakes is not inherently malicious. They are the same AI techniques that power helpful innovations like voice assistants, language translation, and medical imaging. However, when directed toward deception, these tools can cause significant harm. If traditional Photoshop manipulation was like carefully painting changes onto a canvas, deepfakes are more like teaching a computer to become an expert forger who can replicate an entire artistic style. The AI doesn't just modify pixels but understand patterns and can generate new content that matches those patterns.

Why deepfakes matter now more than ever

The democratization of deep-fake technology presents unprecedented challenges. What once required extensive technical knowledge and computing resources now become user-friendly apps accessible to anyone with a consumer device (smartphone, laptops, etc.) For example, FaceApp, a popular face-swapping app that was downloaded over 100 million times before facing scrutiny over privacy concerns as the user data are being stored without consent. This surge in deep-fake apps and accessibility gradually worsens digital literacy as more people struggle to identify misinformation. As a result, the concept of verifiable truth suffers when authentic content can be dismissed as “fake”, and fake content can appear authentic. Besides, targeted manipulation becomes a problem as deepfakes disproportionately target women through non-consensual intimate imagery, and increasingly, political figures through disinformation campaigns.

Research from the University of Amsterdam found that even media professionals struggle to distinguish deepfakes from authentic content. If experts can be fooled, what chance does the average user have?

The "liar's dividend" and deep-fake’s damage

Perhaps the most insidious effect of deepfakes isn't the fakes themselves, but what researchers call the "liar's dividend", the ability of wrongdoers to dismiss authentic evidence as fake.

When former President Trump's lawyers questioned the authenticity of the infamous "Access Hollywood" recording in which he boasted about grabbing women without consent, they were leveraging this concept. As deepfakes proliferate, this defense becomes increasingly plausible to the public.

"Anything can be fake; nothing has to be real," explains Dr. Hany Farid, a digital forensics expert at UC Berkeley. This dynamic creates a perfect tool for accountability denial, especially for powerful individuals and institutions.

While political deepfakes capture headlines, the technology's impact extends far beyond. In 2019, an audio deepfake successfully impersonated a CEO's voice to authorize a fraudulent transfer of €220,000. Meanwhile, deepfake pornography accounts for an estimated 96% of all deepfake videos online, with virtually all victims being women.

Victims of deepfake pornography face severe psychological impacts. According to a study across the UK, New Zealand, and Australia by Flynn et al. (2022), victims experience "a range of emotional, psychological, occupational, and relational effects, many of which continued long after the abuse had first taken place." As one researcher in the MIT Technology Review noted, the toll on victims can be devastating, some have "had to change their names" while others have "completely removed themselves from the internet" out of fear that the images could resurface and "once again ruin their lives."

Technical solutions and their limitations

The tech industry has responded with various detection tools, software designed to identify deepfake signatures. For example, Microsoft's Video Authenticator and Intel's FakeCatcher attempts to combat digital misinformation by analyzing subtle signs that human perception might miss.

However, we're locked in an arms race. As detection technology improves, so do the deepfakes themselves. Many studies show detection accuracy declining as deep-fake generators evolve to counter known detection methods.

Technical detection tools face significant limitations in this arms race. Claire Wardle of First Draft, a nonprofit focused on digital misinformation and co-author of the foundational report "Information Disorder: An interdisciplinary Framework for Research and Policy," advocates for taking approaches that combine technological tools with media literacy and policy frameworks rather than relying on technical solutions alone.

My opinion

To combat digital misinformation, I believe our response must focus on building digital resilience rather than chasing perfect technical solutions or implementing overly restrictive regulations that might slow down innovation. If deepfakes are the new normal, then conscious media consumption must become a core skill taught as part of formal education and employee training. As an example, Finland's successful nationwide media literacy initiative shows a promising model offering remarkable resilience against misinformation campaigns. Furthermore, rather than just detecting fakes, we should focus on verifying authenticity. The Content Authenticity Initiative, with members including Adobe, Twitter, and The New York Times, is developing open standards for content provenance. This "nutrition label" approach would create technical mechanisms to track digital content from the moment it is created through publication.

While some jurisdictions have enacted deepfake-specific legislation, these efforts often focus narrowly on political or pornographic content. Despite this being a positive step towards fighting against deep-fake, we need comprehensive frameworks that address technology’s range of harms while protecting legitimate uses in art, education, and entertainment.

Moving forward

Ultimately, confronting deep-fakes requires recognizing our shared responsibility for the information ecosystem. Just as we wouldn't dump toxic waste into nature, we must develop approaches against polluting our collective digital environment. This means rethinking the standards we use to determine what information deserves our trust and the standard for authenticity.

The challenge of deepfakes isn't just technological but also social and psychological. The question isn't whether we can perfectly detect every deepfake, but whether we can maintain common understanding on the standards of what real information is. As AI continues its rapid evolution, our response to deepfakes will shape not just how we consume media, but how we function as a society.

This article is licensed under CC BY-NC-SA 4.0