AI-generated image-based sexual abuse: prevalence, attitudes, and impacts

The research also examines the ethical, legal and policy implications of AI-generated image-based sexual abuse, and people’s attitudes, opinions, and perspectives on this topic.

This research project on artificial intelligence and image-based sexual abuse (AI-IBSA) is a collaboration between RMIT University, Deakin University, Queensland University of Technology, and the Google Trust & Safety Research team. The project aims to:

1. Investigate the general public’s attitudes towards AI-IBSA;
2. Empirically measure and examine the prevalence of AI-IBSA perpetration and victimization;
3. Investigate how different forms of AI (e.g., intimate image generators) are being used, or could be used, to perpetrate AI-IBSA; and
4. Explore the digital tools or approaches that societal actors are using, or could use, to detect, prevent and respond to AI-IBSA.

Through a desk-based review, interviews with stakeholders, and a multi-country online survey, this project will generate much needed empirical evidence on AI-IBSA. The expected benefits of this project include increased understandings of how different types of AI technologies are being, or could be, used to perpetrate AI-IBSA; an understanding of how digital tools are being, or could be, used to better detect, prevent and respond to this type of abuse in practice; enhanced industry and scholarly collaborations; research outputs; and recommendations for research, policy, and practice.

Please note:

Image-based sexual abuse (IBSA) refers to the non-consensual taking, creating, or sharing of intimate (nude or sexual) images, including threats to share intimate images.

‘Generative AI’ is a form of AI which refers to algorithms which are trained on existing ‘data’ (e.g., books, web text, photos, videos) to create new, high-quality content, such as new text, photos, or videos.

Deepfakes refers to the creation of realistic-looking but fake photos or videos using AI to depict a person doing or saying something that they haven’t said or done. These images are created by superimposing or `stitching' a person's face onto another another person's body, or by altering voice, facial expressions, or body movements of a person in a video.

• Nicola Henry
• Lisa Given
• Alexa Ridgway
• Gemma Beard

2024

Google

Thumbnail image credit: Alexandra Koch (https://pixabay.com/photos/ai-robot-artificial-intelligence-7977960/)