Other
A scoping review of threats, enabling technologies, and forms of AI involvement associated with child sexual exploitation and abuse
Chad Lance Hemady, Sara Valdebenito, Chanel Hayre, Franziska Meinck, Deborah Ann Fry ·
Summary
Written by AI, not yet AICS's view · claude-haiku-4-5-20251001This scoping review examined how artificial intelligence systems are being used in child sexual exploitation and abuse (CSEA). The review identified five main threat domains: AI-generated child sexual abuse material, AI-manipulated material, grooming, sextortion, and conversational AI sexual abuse. The authors found that AI is reshaping established forms of abuse through new production methods and forms of interaction, and they propose a framework for classifying these harms that goes beyond existing violence classifications.
Key points
- Five threat domains identified: AI-generated CSAM, AI-manipulated CSAM, grooming, sextortion, and conversational AI sexual abuse
- Six technology groups enable CSEA: generative systems, model adaptation and training data, media manipulation systems, conversational systems, immersive technologies, and access/hosting infrastructure
- Three forms of AI involvement: AI-enabled assistance with abuse, AI-simulated co-production of scenarios, and AI-driven initiation or escalation of sexualised conduct
- Inconsistent terminology has previously limited understanding of AI-CSEA threats
- AI-CSEA largely reflects established abuse forms reshaped by new production and interaction modes
Limitations
This is a scoping review synthesising existing literature rather than primary research, so it cannot evaluate the prevalence or scale of these harms or provide empirical evidence of intervention effectiveness. The review is limited to English-language sources, which may not capture global perspectives on AI-CSEA threats.
Abstract
Artificial intelligence (AI) is increasingly implicated in child sexual exploitation and abuse (CSEA), but inconsistent terminology has limited synthesis. This scoping review mapped reported AI-CSEA threats and enabling technologies and examined how AI systems participated in perpetration. English-language sources published from January 2010 to 1 April 2026 were eligible if they concerned children under 18 and where an AI system played a material role in the abuse process. Searches of the ACM Digital Library, CINAHL, Web of Science, EBSCOhost databases, and grey literature identified 48 sources. Three reviewers contributed to screening and data charting, and extracted data were summarised numerically and through qualitative content analysis. Overall, our review found that AI-CSEA nomenclature conflated abusive conduct, production methods and system roles. Five non-exclusive threat domains comprised i) AI-generated CSAM, ii) AI-manipulated CSAM, iii) grooming, iv) sextortion, and v) conversational AI sexual abuse. Six technology groups encompassed: i) generative systems; ii) model adaptation and training data; iii) media manipulation systems; iv) conversational systems; v) immersive and embodied technologies; and vi) access, hosting, and distribution infrastructure. Forms of AI involvement include: i) AI-enabled assistance with person-directed abuse; ii) AI-simulated co-production of user-initiated scenarios; and iii) AI-driven initiation or escalation of sexualised conduct. AI-CSEA largely reflected established forms of abuse reshaped by new modes of production and interaction. Existing act-based classifications of violence against children provide a suitable foundation but require supplementary recording of production route, child linkage, system role and point of steering.