AI-Generated Child Sexual Abuse Material: Technology, Legislation, and a Call for School Action

By Neleh Kay, Brown University, Class of 2027
As a current college student, I’ve become acutely aware of how generative artificial intelligence (AI) is transforming education: my peers refer to ChatGPT by nickname, my professors constantly debate its place in classrooms, and my education studies courses consider its potential to address educational equity issues. But while I had thought at length about the academic implications of AI, I was not yet familiar with its nefarious use in child sexual abuse material (CSAM) production.
In 2025 alone, a staggering 145,000 incidents of AI CSAM were reported, that is, CSAM that is either generated or altered by AI software. However jarring or unsettling this may be, we need to confront these realities head-on. The current AI boom shows no signs of slowing down, and such cases are likely just the start of a ballooning crisis.
Background: A Pre-AI Digital Era
To better understand AI CSAM, we have to understand what came before: before ChatGPT, Google Gemini, and the assortment of “nudify” apps highly accessible today. CSAM is not new. The pre-AI cyberspace was an incubator for criminal activity, especially child sexual abuse.
For one, digital content has long carried the risk of proliferation. Images, whether sexually explicit or not, can be multiplied at lightning speed, across countless platforms. A 2019 article reported that CSAM is found on major platforms such as Facebook Messenger, Dropbox, Tumblr, and Snapchat. More covert are the dark web forums, estimated to have millions of photos and videos shared between tens of thousands of members. The darknet promises users anonymity via encryption, making it nearly impossible to detect and remove CSAM from its websites.
Other shortcomings lie in the procedure following CSAM discovery. Historically, major tech companies have failed to cooperate with law enforcement and hotline workers, becoming complicit in the proliferation and trade of CSAM. Findings from 2020 reveal that, upon CSAM detection, popular websites terminated user accounts or notified account holders of ongoing investigations. Both protocols hinder retrieval efforts and ultimately, the processes of establishing culpability and delivering justice.
More complications arise with the National Center for Missing and Exploited Children, the organization responsible for distributing reports to local, state, and federal law enforcement agencies. Because of an overwhelming volume of CSAM reports, the nonprofit center has been forced to focus on those cases concerning infants and toddlers, leaving older children and teenagers without immediate attention.
Fast-forward to today, and these conditions haven’t vanished–if anything, they’ve worsened. Accordingly, it’s imperative that we not isolate AI CSAM from the online infrastructure and vulnerabilities that gave rise to it.
A New Frontier: AI CSAM
First and foremost, it is important to know exactly what AI CSAM refers to. In line with the vast array of public-facing AI applications, AI CSAM has taken many forms. The most common of which are deepfakes: images, videos, or audio that lead viewers to believe someone has said or done something they did not actually say or do. It might look like photos and videos of a child’s face overlaid onto an adult body, or ones of their own body “nudified” to appear as if without clothing. The three most consistently abused AI models are large language models (LLMs), text-to-image models, and “nudify” apps. “Nudify” apps are particularly dangerous because they’re trained on massive data collections of pornographic imagery and require virtually zero technical skills to operate.
Across international reports, key characteristics and patterns of AI CSAM have been observed. With regard to gender, girls are disproportionately victimized. Last year, girls were depicted in 97% of images where simulated sex was recorded. This gender disparity goes beyond bad actors to the biases embedded within the algorithms themselves. One case study on DeepNude, an early “nudify” app, demonstrated that its algorithm was fundamentally incapable of performing its functions on images of men. The software was trained on images of women and women only, engineered with the purpose of sexualizing female bodies. Findings like these remind us that AI CSAM does not exist in a vacuum, but as a product of the decades-long history of cyberviolence against women and girls that began with cyberstalking and gender-based hate speech, and persists with the misogynistic rhetoric of the “manosphere.”
Another defining characteristic of AI CSAM is that it is commonly produced by minors, marking a shift in traditional conceptualizations of CSAM, which are typically characterized by dynamics between adult offenders and young children. Cases of male students using AI to make CSAM of their female peers have sprung up nationwide in at least eight states. According to the National Education Association, administrators and educators are not yet equipped to handle student-on-student incidents. They may be unsure whether to report AI CSAM, as it isn’t always acknowledged in definitions of child abuse. Until now, the existence and enforcement of accountability measures for student offenders have been murky. One high school student was ordered to complete treatment for creating 43 AI images and videos of his underage classmates, but most of the charges against him were dropped because relevant state law did not yet exist. Inconsistency in state legislation and school policy, and delays in updating both, are likely to perpetuate these shortcomings in accountability for future juvenile perpetrators.
Next, we will consider how and why AI CSAM causes harm. The 2026 AI CSAM Report, conducted by the Internet Watch Foundation, highlights some important dimensions of harm felt by victims, their support systems, and society at large: revictimization and normalization of sexual violence against children. Survivors of AI CSAM experience revictimization when their original abuse images are generated anew–frequently into more extreme versions–or when imagery is circulated across group chats, forums, and public image galleries for years to come. Repeated trauma of this kind causes serious and lasting psychological damage, including anxiety, depression, and social withdrawal.
The second concern, normalization of child sexual violence, is a societal consequence of the widespread dissemination of AI CSAM. Research has shown that CSAM consumption reinforces users’ sexual interests in children and may lead them to seek out interaction with real children, outcomes that are exacerbated by repeated exposure. Knowing this, AI CSAM raises concerns about the potential pathways between online and offline forms of abuse.
The Current State of Legislation and School Policy
Given the novelty of AI CSAM and deepfake pornography, and the speed at which the underlying technology is advancing, the law has been slow to catch up. In the United States, both federal and state law remains fragmented at best, enabling offenders to exploit weaknesses and loopholes.
One landmark law, the TAKE IT DOWN Act, was signed into law on May 19, 2025. It made history as the first piece of federal legislation to meaningfully address the spread of non-consensual intimate imagery (NCII), which includes AI-generated deepfake pornography that features identifiable individuals. On the matter of AI CSAM, the TAKE IT DOWN Act criminalizes offenders who knowingly publish actual or synthetic CSAM, and those who intentionally threaten to publish such material in order to coerce, extort, or intimidate victims. Additionally, TAKE IT DOWN holds major tech companies accountable by mandating removal of NCII and CSAM from online platforms within 48 hours of a survivor’s request. So, despite not going as far as to disallow “nudify” apps altogether, it marks a historic first step of federal legislative efforts to combat AI CSAM.
Alongside TAKE IT DOWN, numerous other federal legislative proposals have been introduced and are awaiting passage. The Stop Sextortion Act would criminalize sextortion or the threatening to distribute CSAM to intimidate, coerce, extort, or cause emotional distress, as well as expanding obscenity laws to include AI CSAM. The SAFE Act would update current CSAM guidelines to better account for modern technology, and the NO FAKES Act would prohibit the creation and distribution of unauthorized digital replicas or digital content made using generative AI.
That said, the bulk of legislative progress has occurred at the state level: 45 states have enacted laws that criminalize AI and computer-edited CSAM, though the specific language varies from statute to statute. Many states only criminalize AI CSAM when a real, identifiable child is depicted. Others criminalize AI CSAM when any person suspected to be a minor is depicted, identifiable or not.
Although most AI CSAM originates beyond school walls, schools have frequently been at the center of this issue. Even so, a regional study found that few districts have established any cyberbullying policies that include mention of online sexual misconduct. If schools and their leadership are to protect students, support survivors, and prevent further harm, they need to set clear policies and procedures for AI CSAM response.
Recommendations for School Response
1) Schools should stay informed about the ever-evolving laws governing AI CSAM in their state.
2) Schools should preemptively set and update policies and procedures to cover AI CSAM, including but not limited to student-on-student incidents, safe reporting mechanisms, and deepfake detection tools.
3) Schools should provide students, teachers, faculty, and parents with education around critical media literacy, AI CSAM, and online sexual misconduct.
4) Schools should allocate resources to trauma-informed mental health and support programs targeted for survivors of AI CSAM.



