This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A Wisconsin man was arrested by the FBI for allegedly using the open-source AI model Stable Diffusion to generate approximately 13,000 sexually explicit images of minors. The perpetrator distributed these images via Telegram and Instagram, including sending them directly to a 15-year-old boy. This case is noted as one of the first federal prosecutions involving the use of generative AI to create child sexual abuse material (CSAM).
The FBI has arrested Steven Anderegg of Holmen, Wisconsin for having allegedly used Stable Diffusion to generate about 13,000 sexually explicit images of minors, which he then is also alleged to have shared and distributed, including with at least one minor, via Telegram and Instagram. Anderegg was originally apprehended by state police in March, and this case marks one of the first times the FBI has brought charges against someone for having used AI to generate CSAM.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI that exposes users to harmful, abusive, unsafe or inappropriate content. May involve providing advice or encouraging action. Examples of toxic content include hate speech, violence, extremism, illegal acts, or child sexual abuse material, as well as content that violates community norms such as profanity, inflammatory political speech, or pornography.
Also classified as
The perpetrator used the generated images and direct messages to communicate with and potentially groom a 15-year-old boy.
The perpetrator used Stable Diffusion to generate and distribute child sexual abuse material (CSAM), which is explicitly classified under exposure to toxic content.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The incident was caused by a human user intentionally utilizing a deployed AI model (Stable Diffusion) to generate child sexual abuse material.
Added categories peak at 4: Severe (Child Sexual Exploitation & Abuse (CSEA), Direct)
An individual used an open-source generative AI model to produce and distribute child sexual abuse material. While representing a severe domestic crime and a novel legal challenge for law enforcement, the incident has minimal direct implications for broader national security, critical infrastructure, or state sovereignty.
The report describes the AI system as a text-to-image generative model used by the perpetrator who typed specific text prompts to generate images. It operated as a tool, performing a specific function on direct input without autonomous decision-making.
Unacceptable Risk: The report describes the use of AI to generate child sexual abuse material, which represents an exploitative system that violates fundamental human rights and poses significant dangers to society.
The AI system, Stable Diffusion, was used to generate hyper-realistic images of minors from text prompts.
The reports are highly detailed, originating from credible sources (The Guardian, Washington Post, Department of Justice), and clearly outline the perpetrator's actions, the AI model used, and the legal charges.
None
Technology
Open-Source Offline Execution
Allows running models locally without cloud-based safety filters.
Flawed Prompt Filtering Bypass
Negative prompts allowed user to bypass adult-generation requirements.
Lack of Local Safeguards
Earlier versions lacked hardcoded blocking of explicit prompts locally.
Data Inputs
CSAM in Training Datasets
LAION-5B database used to train models contained child exploitation images.
Human Factors
Malicious User Intent
User actively generated, stored, and distributed explicit images of minors.
Exploitation of Minors
User sent explicit AI images to a minor to groom and communicate.
Online Offender Communities
Predators used social media to share prompts and validate interests.
Process and Methods
Inadequate Dataset Auditing
Failure to clean training datasets of illegal content prior to release.
Lack of Local Verification
No mechanism to verify or report locally-run generation activities.
Management
Premature Model Release
Releasing open-source models before robust safety filters were integrated.
Safety Safeguard Delegation
Split model development across entities led to fragmented safety oversight.
Regulatory Environment
Evolving Legal Enforcement
First-of-its-kind application of federal CSAM laws to fully AI-generated media.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
