1.

Introduction
Artificial intelligence has transformed nearly every facet of human life, from healthcare and
education to art and entertainment. Yet, in parallel with its innovation, AI has also become a
mechanism for exploitation. Deepfake pornography—the creation of synthetic sexually
explicit videos or images featuring the likeness of a person without their consent—represents
one of the most disturbing outcomes of AI misuse.

Unlike conventional image manipulation, deepfakes use advanced machine learning

algorithms that replicate a person’s facial expressions, voice, and movements with alarming
realism. In the hands of malicious actors, this technology is primarily weaponized against
women. The global implications are staggering: according to a 2024 UN Women report, 95%
of all deepfake videos on the internet are pornographic, and over 98% of them feature
women. Victims range from celebrities and journalists to private citizens who have never
sought public attention.

Deepfake pornography does not merely infringe upon privacy; it violates bodily integrity,
digital identity, and autonomy. The victims’ images are transformed into tools of
humiliation and control. This paper seeks to analyze the phenomenon in its entirety—
technological, sociological, psychological, and legal—arguing that deepfake pornography
must be recognized globally as a human rights violation rather than a mere technological
misuse.

2. The Technology Behind Deepfakes

Deepfake creation relies on Generative Adversarial Networks (GANs), a class of AI
algorithms introduced by Ian Goodfellow in 2014. GANs consist of two neural networks — a
generator that creates synthetic data and a discriminator that evaluates it. Through iterative
competition, the generator becomes capable of producing highly realistic images or videos
that can deceive both humans and automated detection systems.

Originally, GANs had noble purposes: generating synthetic data for training other AI models,
restoring damaged images, or simulating virtual environments for education and design.
However, the democratization of these tools led to their misuse. By 2022, open-source
software such as DeepFaceLab, FaceSwap, and DeepFake Web enabled users with minimal
technical skill to create convincing manipulations.

A 2023 study by the University of Cambridge’s Centre for the Study of AI Ethics identified
that most online deepfake generation requests were sexually explicit in nature, and more
than 90% were aimed at non-consensual content. The rise of free diffusion models like
Stable Diffusion and Midjourney only accelerated the accessibility of these tools.

These models learn from enormous datasets—often scraped without consent from social
media platforms or entertainment databases. As a result, women’s digital images, uploaded
innocently to platforms like Instagram or TikTok, become raw material for synthetic
pornography. This exploitation marks a new dimension of violation: it is not merely data
misuse, but the digital colonization of female identity.
3. Gendered Dimensions of Deepfake Exploitation
Deepfake pornography is a distinctly gendered phenomenon. Although anyone can
theoretically be targeted, the overwhelming majority of victims are women. This gender
disparity mirrors broader societal patterns of online abuse, where misogyny and sexual
objectification thrive.

A 2023 Amnesty International report highlights that women — particularly those in the
public eye — are targeted not simply for visibility but for punishment. Female journalists,
politicians, and activists who speak on feminist or political issues face orchestrated deepfake
campaigns designed to discredit them. These fabrications circulate widely before any
verification can occur, undermining credibility and silencing dissent.

For instance, in India’s 2022 Delhi elections, female candidates were targeted with fake
explicit videos that went viral on WhatsApp groups, resulting in withdrawal from public
appearances. Similarly, in South Korea, the infamous “Nth Room” scandal revealed a
massive online community that shared and sold non-consensual sexual content, including
deepfakes of ordinary women.

The phenomenon illustrates that deepfakes are not merely technological pranks—they are
continuations of patriarchal control, digitized. In patriarchal cultures, women’s reputations
are tied to notions of purity and honor; thus, even fabricated sexual imagery can destroy
social standing. The exploitation of deepfake technology amplifies these gendered
vulnerabilities, transforming AI into a mechanism of systemic gender oppression.

4. The Psychological Impact on Victims

Victims of deepfake pornography experience trauma that rivals or exceeds that of other forms
of sexual violence. While no physical act has occurred, the psychological and social
consequences are devastating.

A 2024 study by the University of Oxford found that 72% of deepfake victims suffer from
severe anxiety and depressive symptoms, while 41% report suicidal ideation. Many
describe the experience as “digital rape” — a loss of bodily autonomy and control over one’s
identity.

One victim from the United Kingdom, whose face was superimposed onto explicit videos
viewed millions of times, described the experience:

“It felt like being assaulted in every corner of the internet. Even if people knew it wasn’t real,
they looked at me differently. It never goes away.”

The permanence of digital media compounds the trauma. Even after content removal,
duplicates and backups often persist across mirrored websites, pornographic forums, and the
dark web. Victims face a digital haunting — a lifelong association between their name and
falsified sexual imagery.

Moreover, traditional legal and psychological support systems are ill-equipped to address this
form of abuse. Victims report that police often dismiss complaints due to “lack of physical
evidence” or “technological complexity.” Such responses deepen the trauma, leaving victims
feeling helpless and invisible.

5. Legal Frameworks and the Global Policy Vacuum

Legal systems worldwide are struggling to keep pace with the speed of AI innovation. While
some jurisdictions have begun to legislate against deepfakes, global regulation remains
fragmented and insufficient.

 United States: Only a handful of states — California, Texas, Virginia, and New York
— have enacted laws criminalizing non-consensual deepfakes. Federal law lacks a
comprehensive framework, leaving victims without uniform protection.
 United Kingdom: The Online Safety Act (2023) made sharing deepfake
pornography illegal, but enforcement remains inconsistent, and many victims face
procedural delays.
 South Korea: Following the “Nth Room” case, South Korea introduced stringent
cyber sexual crime laws in 2022, making the creation or distribution of deepfake porn
punishable by up to seven years in prison.
 European Union: The EU AI Act (2024) includes transparency requirements for
synthetic media but does not specifically address the non-consensual pornographic
use of AI.

The international gap in legal response underscores a deeper ethical issue — the global
internet lacks jurisdictional coherence. Deepfakes are created in one country, hosted in
another, and distributed globally, making prosecution extremely difficult.

In many developing nations, where digital literacy and resources are limited, victims face
insurmountable barriers to justice. The lack of cybercrime infrastructure allows perpetrators
to operate with impunity. As a result, deepfake pornography thrives in the grey zones of
international law.