Holmen, Wisconsin Man Allegedly Used Stable Diffusion to Create and Then Share Sexually Explicit Images Depicting Minors

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.

Risk classification

• Primary risk domain: 1 Discrimination & Toxicity
• Primary risk subdomain: 1.2 Exposure to toxic content

The perpetrator used Stable Diffusion to generate and distribute child sexual abuse material (CSAM), which is explicitly classified under exposure to toxic content.

Additional risk subdomains

• 4.3 Fraud, scams, and targeted manipulation: The perpetrator used the generated images and direct messages to communicate with and potentially groom a 15-year-old boy.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

The incident was caused by a human user intentionally utilizing a deployed AI model (Stable Diffusion) to generate child sexual abuse material.

EU AI Act risk tier

• Risk tier: 1 Unacceptable

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.

AI system and alleged parties

• AI system: None named
• AI purpose: Image Generation; Deepfake Image Generation
• Behaviour type: Tool
• Alleged developer: Stable Diffusion, Stability AI
• Alleged deployer: Steven Anderegg
• Alleged harmed parties: minors, General public

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Severe. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Minor, indirect Minor
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Minor, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Minor, indirect Negligible
• Epistemic: direct Minor, indirect Substantial
• Child sexual exploitation and abuse: direct Severe, indirect Severe

Malicious content

Reported: The report explicitly describes toxic and malicious content created and spread directly and indirectly by the incident.

Directly caused: The perpetrator generated approximately 13,000 hyper-realistic images of nude and semi-clothed prepubescent children using Stable Diffusion.

Indirectly caused: The images were distributed to other offenders on Telegram and Instagram, spreading toxic content.

Inferred additional harm: Widespread proliferation of similar AI-generated CSAM across online forums and dark web networks.

Civil rights

Reported: The report explicitly describes violations of human and civil rights caused directly by the incident.

Directly caused: The generation and distribution of child sexual abuse material violates the fundamental rights of children to be protected from sexual exploitation.

Inferred additional harm: Widespread erosion of children's rights due to the ease of generating synthetic CSAM.

Psychological

Reported: The report explicitly describes psychological or mental-health harm to a minor caused directly by the incident.

Directly caused: A 15-year-old boy was sent explicit AI-generated images of minors, causing potential psychological distress and exposure to abusive material.

Inferred additional harm: Widespread psychological harm to minors targeted by online offender networks, and potential trauma to the 15-year-old victim.

Epistemic

Reported: The report explicitly describes epistemic harm caused directly and indirectly by the incident.

Directly caused: The generation of 13,000 photorealistic fake images of minors.

Indirectly caused: Law enforcement and child safety organizations face severe challenges distinguishing between real and AI-generated CSAM, threatening to overwhelm their systems.

Child sexual exploitation and abuse

Reported: The report explicitly describes a CSEA incident involving AI-generated CSAM and distribution to a minor.

Directly caused: The perpetrator generated 13,000 CSAM images and sent several to a 15-year-old boy.

Indirectly caused: The perpetrator used these images to cultivate an online community of offenders, fueling interest in child sexual abuse.

Inferred additional harm: Widespread facilitation of child grooming and abuse material generation by other bad actors using similar open-source models.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 1

Potential causes

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.

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.

Regulatory Environment

• Evolving Legal Enforcement: First-of-its-kind application of federal CSAM laws to fully AI-generated media.

Information quality

• Classification confidence: High
• Reason for confidence: 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.
• Ambiguities identified: None of significance regarding the primary incident, though there are minor differences in the exact version of Stable Diffusion used.

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.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: An individual US citizen

Threat characteristics

• Imminence: Long-term. The immediate threat has been neutralized by the perpetrator's arrest, representing an ongoing strategic concern regarding AI-enabled illicit content generation.
• Autonomy: Human-controlled. The perpetrator manually operated the AI tool using highly specific text prompts and negative prompts to generate the desired images.
• Novelty: First-of-its-kind. This represents one of the first federal prosecutions involving child sexual abuse material generated entirely by artificial intelligence, setting a new legal precedent.

Impact by dimension

• Physical security: Negligible. The incident involved the generation of digital images and did not pose any threat to physical systems, critical infrastructure, or kinetic safety.
• Information security: Negligible. There is no indication of intelligence compromise, state-sponsored information warfare, or systematic disinformation targeting national institutions.
• Sovereignty: Negligible. No foreign AI systems or actors compromised government decision-making, electoral systems, or core state authority.
• Economic security: Negligible. The incident does not involve strategic technology theft, financial system attacks, or disruption of critical economic supply chains.
• Societal stability: Minor. While representing a severe violation of child safety and human rights, the incident was localized and effectively managed within standard domestic law enforcement procedures.