[論文レビュー] How to Stop Playing Whack-a-Mole: Mapping the Ecosystem of Technologies Facilitating AI-Generated Non-Consensual Intimate Images

Michelle L. Ding, Harini Suresh+1 more|arXiv (Cornell University)|Feb 4, 2026

Gender, Feminism, and Media被引用数 0

ひとこと要約

The paper builds the first comprehensive map of the AIG-NCII technology ecosystem, identifying 11 categories across creation, distribution, proliferation, infrastructure, and monetization, and demonstrates its use via case studies including Grok.

ABSTRACT

The last decade has witnessed a rapid advancement of generative AI technology that significantly scaled the accessibility of AI-generated non-consensual intimate images (AIG-NCII), a form of image-based sexual abuse that disproportionately harms women and girls. There is a patchwork of commendable efforts across industry, policy, academia, and civil society to address AIG-NCII. However, these efforts lack a shared, consistent mental model that situates the technologies they target within the context of a large, interconnected, and ever-evolving technological ecosystem. As a result, interventions remain siloed and are difficult to evaluate and compare, leading to a reactive cycle of whack-a-mole. We contribute the first comprehensive AIG-NCII technological ecosystem that maps and taxonomizes 11 categories of technologies facilitating the creation, distribution, proliferation and discovery, infrastructural support, and monetization of AIG-NCII. First, we build and visualize the ecosystem through a synthesis of over a hundred primary sources from researchers, journalists, advocates, policymakers, and technologists. Next, we demonstrate how stakeholders can use the ecosystem as a tool to 1) understand new incidents of harm via a case study of Grok and 2) evaluate existing interventions via three more case studies. We conclude with three actionable recommendations, namely that stakeholders should 1) use the ecosystem to map out state, federal, and international laws to produce a clearer policy landscape, 2) collectively develop a database that dynamically tracks the 11 technologies in the ecosystem to better evaluate interventions, and 3) adopt a relational approach to researching AIG-NCII to better understand how the ecosystem technologies interact.

研究の動機と目的

• Synthesize a wide range of sources to map the technologies enabling AIG-NCII.
• Provide a shared framework and terminology for stakeholders across policy, industry, and civil society.
• Demonstrate how the ecosystem aids understanding of harm incidents and evaluation of interventions.
• Offer actionable recommendations to improve policy, data tracking, and relational research approaches.

提案手法

• Perform a literature and artifact synthesis from >100 sources across academia, journalism, advocacy, policy, and industry.
• Develop a taxonomy of 11 technologies grouped by role (creation, distribution, proliferation/discovery, infrastructure, monetization).
• Triangulate claims with technical artifacts, terms of service, and platform documentation; conduct limited primary checks (e.g., test accounts) to validate ecosystem roles.
• Visualize the ecosystem with a map (Figure 1) and descriptive case studies.
• Iteratively fill gaps by seeking additional sources on developer platforms and infrastructure to maintain granularity and generalizability.

Figure 1. The ecosystem of technologies that facilitate AI-generated non-consensual intimate images (AIG-NCII) via A. Creation (1. training datasets, 2. generative AI models, and 3. generative AI interfaces), B. Distribution (4. distribution channels), C. Proliferation & Discovery (5. deepfake creat

実験結果

リサーチクエスチョン

• RQ1 What are the technological components that collectively enable AIG-NCII creation, distribution, and monetization?
• RQ2 How can a unified ecosystem map improve understanding of new AIG-NCII incidents and inform intervention design?
• RQ3 What actions (policy, data infrastructure, relational research) best reduce reactive, whack-a-mole responses to AIG-NCII?

主な発見

• Identifies 11 technology categories spanning creation, distribution, proliferation/discovery, infrastructure, and monetization.
• Provides a visual ecosystem map (Figure 1) and descriptions with historical context for each category.
• Uses Grok as a case study to show how the ecosystem explains harm and the limitations of platform responses (e.g., access restrictions).
• Demonstrates how the ecosystem can guide policy landscape mapping across state, federal, and international laws.
• Argues for a centralized, dynamic database tracking the 11 technologies to support evaluation of interventions.
• Advocates a relational research approach focused on the edges of the ecosystem to understand interactions among technologies.

Figure 2. Screenshot of grok.com/plans . Accessed January 11, 2026.