As Domestic Violence Awareness Month comes to a close this week, one University of Florida researcher is examining the roots of nudification apps, online platforms that use AI to turn images of clothed people — easily and convincingly — into nude images.
"There's a strong link between nonconsensual intimate imagery and domestic violence," said Kevin R. B. Butler, Ph.D.
There have been legislative and legal responses to the creation and distribution of these images, which are more correctly known as synthetic non-consensual intimate imagery, or SNCII. In the United States, the TAKE IT DOWN Act was passed in 2025, criminalizing the creation and distribution of such imagery and making it easier to remove once posted.
"Real or synthetic imagery can be used to threaten exposure, known as 'sextortion,'" Butler said. "This type of imagery can be used to prevent a partner from leaving, reporting violence or seeking custody of children because of the perceived reputational harm."
In his latest research project, Butler is examining the problem at its root: the tools and communities that produce these nonconsensual images.
Butler is a professor and chair of the University of Florida Department of Computer & Information Science & Engineering, or CISE. He and a group of CISE graduate students, along with a colleague from Georgetown University, wrote the paper "It Takes a Forum: Resources & Stakeholders Sustaining the Online Synthetic Non-Consensual Intimate Image Creation Ecosystem," which investigates SNCII creation and dissemination from a resource-centric perspective.
By exploring the online forums commonly used by criminal nudifiers, the authors are identifying the resources and knowledge shared by the ad-hoc communities involved in SNCII, as well as the mechanisms that spread the resources and knowledge. For Butler and his co-authors, understanding the underlying tools and systems that support SNCII creation, rather than simply the end result, should create a stronger foundation to combat SNCII creation and dissemination.
The work will be presented in November at the Association for Computing Machinery 2026 Conference on Computer and Communication Security in The Hague, the Netherlands. A pre-print is available prior to the conference.
Butler's co-authors were UF doctoral students Bernardo B. P. Medeiros, Malvika Jadhav and Allison Lu, CISE faculty member Vincent Bindschaedler, Ph.D., and Tadayoshi Kohno, Ph.D., from Georgetown University.
The authors found that many malicious actors responsible for disseminating SNCII operate within internet forums to exchange resources, strategies and generated content across multiple platforms. The results indicate that technically sophisticated actors gravitated toward certain communities (anonymous bulletin board 4chan, for example), while less tech-savvy end users are more active on other platforms such as Reddit.
Researchers analyzed 282,154 4chan posts and 78,308 Reddit submissions from June 9 to Nov. 21, 2025. Through in-depth manual analyses of a random sampling of 200 posts from each platform, they developed a multi-label behavioral codebook of actors, actions and resources used by each dataset. The codebook was then used as the source for analysis of the full datasets.
Further analysis revealed a shadow infrastructure that adapts to enforcement, which facilitates the sharing of model customization resources such as tutorials or base models, through private messaging applications such as Telegram and Session.
Researchers found that individuals called "educators" acted as bridges to help newcomers transition into active SNCII creators. These educators would provide interactive troubleshooting, resource curation and follow-up guidance; the team reports that 56.3 percent of these educators helped transition less experienced users into more advanced users. By analyzing patterns in the data, Butler's team was able to zero in on these educators.
More worrisome was what the researchers described as an accountability gap among nudifiers. Many users displayed familiarity with regulations prohibiting their activities but were unconcerned about the effects on their SNCII creation and sharing activities.
"The fact that they were so open about their sharing of resources and images was one of the things that motivated us to begin this study," said Medeiros, CISE doctoral student and co-lead author of the paper. "We began this work soon after the TAKE IT DOWN act, but despite this, we found that the 4chan community was not deterred from posting images. Even today, people are still posting these kinds of images on 4chan."
The researchers were able to identify some possible intervention strategies to stem the tide. Having developed ways in which to identify nudification educators, the team could mitigate influencers' influence in a focused, targeted manner.
Another possible chokepoint suggested in the paper focused on major public repositories — photo and file storage platforms — of user-trained AI image-generation models. The researchers suggest that companies could enact meaningful prohibitive policies, the impact of which could easily be measured.
It should be noted that the researchers were very mindful of the ethical considerations involved in this type of research. Researchers scraped data from Reddit and 4chan forums but never directly interacted with or contributed to the traffic of any illicit forum communities. Data was anonymized and no imagery was viewed or saved.
For Butler and the team, the work's impact extends much further than the realms of computer science. Ultimately, the hope is to identify gaps in the existing regulatory landscape, provide intervention points to deter the bad guys and, certainly, protect privacy and safety of users and their most sensitive and vulnerable information.
