By Ṣikẹmi Akinmade
As recent outrage over sexually manipulated images disseminated by Elon Musk’s AI tool, Grok, has been driving discourses around AI safety and platform accountability, the perilous connection between generative AI and child sexual exploitation and abuse (CSEA) remains structurally misunderstood.
While harm is often considered only in terms of ‘circulated material deepfakes’, AI has shifted the consumption of non-consensual intimate images (NCII) from the public to the private sphere, transforming one-on-one violence into demand-driven cycles of mass-exploitation. This mutation becomes especially explicit in the rise of organised AI generation, characterised by the production, commercialisation, and even commissioning of child sexual abuse materials (CSAM).
Lowering barriers to structural sexual abuse
AI reproduces existing forms of misogynistic, sexual and gender-based violence, magnifying the sociopolitical concessions we make to issues of abuse in society every day. It is, however, the explicit removal of ethical, legal and practical barriers to CSEA that sets AI-fueled sexual violence apart from previous abuse contexts.
According to Ana Ornelas, Policy Officer for digital rights at the European Sex Workers’ Rights Alliance (ESWA), “generative AI reduces friction when it comes to NCII-sharing, as the technology makes it easier to both create and spread the content”.
No longer limited by what sex and adult content workers can and will legally do, generative AI makes sexual abuse faster, less risky and more accessible for both current and potential abusers, especially when it comes to children.
Dan Sexton, Chief Technology Officer of the Internet Watch Foundation against child sexual abuse (IWF), stresses this, stating that “viewing CSAM, whether photographic or synthetic, normalises and entrenches the sexual abuse of children”.
In a clear escalation of violent potential, the IWF found sexual material generated by AI to have a 20 per cent greater presence than human-made videos in the most extreme category of abuse, with 65 per cent of AI-generated videos containing penetration, torture, sadistic acts and bestiality.
Julie Fuchs, EU Policy and Advocacy Officer at ECPAT, a global network dedicated to eradicating the sexual exploitation of children, explains this as an addiction-like trajectory where “as the initial thrill fades, individuals seek increasingly extreme and violent content”.
The danger then lies not only in the scope of extremely abusive content, but in the inevitable acceleration and scaled expansion of paedophilic abuse. It is this fusing of generative AI capabilities with industrialised exploitative production that renders strictly platform-focused debates structurally inadequate in addressing the complex emerging systems of violence against children.
Criminal networks & the commercialisation of sexual violence
As the sexual content industry increasingly moves towards the use of generative AI tools in their media production and tech companies further venture into selling sexualised services, it is the most vulnerable individuals in society who get caught at the intersection of media and further marginalisation. This is especially concerning, as it incentivises producers of such products to dedicate themselves to creating sexual abuse materials that attract, monetise and amplify the paedophilic demand of others.
On the one hand, this occurs through criminal networks of global AI production that often scale their operations faster than international regulations or enforcement, like Europol, can keep pace.
On the other hand, organised AI production can also occur through individuals who are commissioned to create personalised deepfake media, allowing abusers to cultivate their depraved fantasies online, unbeknownst to their victims – many times, children with whom they have personal relations. Convicted UK resident, Hugh Nelson, for example, stated in his police interview that many of the deepfake child characters he would create sexual abuse material of were commissioned by the victims’ fathers, uncles, and family friends.
What then emerges is a demand economy of abuse that is not accidental, but structural: producers, commissioners and consumers connected through AI-enabled supply chains, critically contributing to a documented increase in pedophilic offline abuse tied to generative AI.
Offline harm, invisibilisation & the myth of ‘victimless crimes’
The surging criminal cases manifest the shift in the consumption of sexualised deepfakes – from primarily appearing on public platforms to being privately exhausted, reinforcing the ongoing invisibilisation of child sexual abuse victims, and highlighting the necessity of considering the data on offline abuse in relation to what is happening online.
ECPAT’s Julie Fuchs warns that exposure to CSAM “often progresses from passive viewing to collecting, connecting with other offenders, and ultimately moving closer to in-person abuse”. She also reflects on research from Finnish-based Protect Children which “found that over half of surveyed child sexual abuse material viewers feared their consumption would lead them to abuse a child in person — and more than one-third said it already had pushed them to seek contact with a child”.
The commercialised mass-production of exploitative materials is, therefore, not only industrialising the abuse, but it is also essentially funding, inspiring and materialising sexual violence with an alarming trauma multiplication effect: The AI tools are constantly being trained to generate based on images of real children whose previous abuse has often already been photographed, tying each new generative request to an infinite pipeline of recycled and repeated child sexual abuse. In the hands of the abusers, the materials are then once again used to coerce, groom and justify the ongoing victimisation and retraumatisation.
This can also be seen in the reignited Epstein file case, where over 180,000 images and 2,000 videos related to sexual abuse and gender-based violence were carelessly released into the online ecosystem. While the widespread minimisation of the disseminated materials wrongfully dismisses much of the depicted abuse as “just AI”, it simultaneously misses that the engagement of both “real” and “synthetic” abuse materials plays a crucial role in furthering the demand for and normalisation of sexual violence.
Fuchs goes on to corroborate this, explaining that although AI “blurs the line between what people think is ‘harmless’ and what leads to in-person harm, there is no safe or harmless version of AI-generated child sexual abuse material. (And that) treating it as a ‘lesser evil’ only enables more offenders, more victims, and more severe abuse”.
Beyond platform regulations – intervention at the root
Both the Grok and Epstein cases exemplify this mismatch by highlighting how public discourses focus on outrage over the dissemination of CSAM, rather than their ongoing creation, as the sustenance of sexual abuse economies and further invisibilisation of affected victims continues to escalate.
ESWA’s Ana Ornelas, echoing the consensus of other humanitarian activists, affirmed that “both the content guidelines of platforms and the regulatory frameworks fail to truly understand the extent of the harm and address it with the victims’ well-being in mind”.
This comes as no surprise, as we are concurrently witnessing the unbridled use of AI in reimagining other forms of contemporary structural violence targeting children and other marginalised groups – from promoting racist, classist and queerphobic hatred to callously illustrating lethal brutality against women.
Addressing this growing multi-sectoral crisis requires a shift from platform-focused debates to establishing a clear survivor-led harm prevention framework and legally backed social incentives that target the systems furthering child sexual abuse globally at the root — in the production, demand and normalisation of violence.
According to Ornelas, “We need stronger mechanisms of redress and protection for victims both in legislation and platforms, but this is not enough. AI-generated NCII-sharing is more of a symptom than a problem in itself. We need better AI literacy so people can truly understand how this content is generated, and we need comprehensive sex education that encompasses the digital sphere.
“Techno solutions can only go so far – we need to address the causes moving people to use this technology to abuse and attack. When it comes to CSAM, we tend to focus on the “Materials” part rather than the “Abuse” – focusing too much on the content that is being spread instead of preventing it from being created in the first place,” she added.
The views and opinions expressed in this article are solely those of the author and do not reflect the official policy or position of the Media Diversity Institute. Any questions or comments should be addressed to the editor at [email protected] of the Media Diversity Institute (MDI).
