# Riana Pfefferkorn: Student Misuse of AI-Powered “Undress” Apps Source: https://www.youtube.com/watch?v=ewS6RacTWGI Recap page: https://rapidrecap.app/video/ewS6RacTWGI Generated: 2025-12-15T17:42:33.299+00:00 --- ## Quick Overview Riana Pfefferkorn's research identifies that student misuse of AI-powered "undress" or "nudify" apps to create non-consensual synthetic imagery (AICSAM/deep fake nudes) is facilitated by easily accessible, purpose-built open-source models, leading to significant psychological harm for victims and exposing gaps in current school preparedness, mandatory reporting laws, and juvenile justice responses. **Key Points:** - The primary source for AICSAM generation involves open-source AI models, often running locally, which are easier to abuse than mainstream hosted models, as their safety guardrails can be removed or were trained on problematic data like Stable Diffusion 1.5. - Purpose-built "nudify" or "undress" apps, underpinning many school incidents, require no technical skill, allowing underage users to create deep fake nudes from a single image, with research finding these apps do not consistently ban image uploads of minors. - Schools are currently unprepared for these incidents, often only addressing online risks reactively, and frequently botch responses by focusing on disciplining the perpetrator rather than supporting victims, which destroys community trust. - Motivations for students creating deep fake nudes range from sheer curiosity and pranks to malicious bullying or use as a form of intimate partner violence, though victims consistently experience extreme psychological harm, including anxiety and fear of lifelong reputational risk. - Legislative responses are evolving; dozens of states recently criminalized AICSAM, but few laws address scenarios where the offender is a minor, leading to skepticism among experts regarding appropriate consequences like criminal prosecution versus counseling. - The federal Take It Down Act criminalizes non-consensual publishing of real or deep fake nudes and mandates platforms remove reported imagery within 48 hours, although this law primarily leaves juvenile prosecution to state jurisdiction. - Schools must prepare by taking anonymous tips seriously, establishing crisis management plans addressing evidence preservation versus victim support (image deletion), and educating students broadly on consent, dignity, and privacy. **Context:** Riana Pfefferkorn, a policy fellow at Stanford's HAI, presented findings from her research with Shelby Gman on the harms caused by generative AI, specifically focusing on AI-created Child Sexual Abuse Material (AICSAM), or 'deep fake nudes,' created by students using AI tools. The presentation contextualized the technology, the evolving legal landscape, the research methodology involving interviews with 52 stakeholders and document review from five public school districts, and concluded with policy proposals for schools, parents, and state lawmakers to address this emerging threat. ## Detailed Analysis Pfefferkorn details that deep fake nudes depicting individuals under 18 qualify as AICSAM, often created by morphing innocuous images of real children into sexually explicit ones, frequently using open-source models like Stable Diffusion 1.5 which underpin purpose-built 'nudify' apps. These apps are easily accessible, require no technical skill, and are lucrative, leading to a 'whack-a-mole' enforcement problem against app stores and advertisers. Legally, while federal law bans morphed imagery, state laws have lagged, though many have recently closed this gap; furthermore, federal law requires platforms to report apparent CSAM to the National Center for Missing and Exploited Children (NCMEC). Research indicates that while the prevalence of these incidents in schools is not yet overwhelmingly high given the 54 million US students, schools are ill-equipped to handle them, often lacking proactive education and crisis plans, sometimes failing to report incidents due to ambiguous mandatory reporter laws or an incentive to 'sweep these incidents under the rug.' Student motivations vary widely, but the resulting trauma is severe, leading to anxiety, school avoidance, and fears of lifelong reputational damage; victims reported feeling 'so out of control' and unable to concentrate. Experts debate consequences, with some favoring counseling and others demanding adult-level criminal prosecution, highlighting that digital interactions seem less consequential to minors. Recommendations include utilizing anonymous tip lines effectively, developing comprehensive response policies, and focusing education on broad principles of consent and dignity rather than specific warnings about bad websites. Lawmakers must clarify mandatory reporter laws—as demonstrated by conflicting cases in Pennsylvania and Washington—and consider defining AICSAM creation as cyberbullying to mandate school policy response without resorting immediately to criminalization, given the documented harm of the juvenile justice system. ### AICSAM Creation Methods - Open-source models like Stable Diffusion 1.5 are the main exponent, often forming the basis for purpose-built 'nudify' apps that enable easy creation of deep fake nudes without technical skill - These apps evade app store review by obscuring functionality and are highly lucrative, estimated to rake in tens of millions of dollars. ### Impact on Victims and School Response - Victims suffer extreme psychological harms, including anxiety and panic attacks, with one noting, 'I was unable to concentrate or feel safe at school' - Schools often botch responses, focusing on the perpetrator over the victim, destroying trust, and sometimes failing to report incidents timely due to legal ambiguity. ### Legal Landscape and Policy Gaps - State laws are catching up to criminalize AICSAM, but few laws address minor offenders, creating a consensus gap on appropriate consequences - The federal Take It Down Act mandates platform removal of reported intimate imagery within 48 hours, but juvenile prosecution remains largely a state matter. ### Recommendations for Schools and Parents - Schools must take anonymous tips seriously, prepare crisis plans balancing evidence preservation with victim support (image deletion), and implement broad education on consent and dignity - Parents must teach children about image respect and ensure they feel safe reporting incidents, while also mitigating risk by limiting their own oversharing of children's images online. ### Juvenile Justice Considerations - Respondents expressed skepticism about criminal charges for minors due to undeveloped cognitive levels, suggesting alternatives like counseling, as seen in South Carolina's misdemeanor approach handled in family court - Researchers note the US jails children at high rates, arguing that the answer to traumatized kids is not traumatizing more kids via the juvenile system.