- Browse/Search the Program
Search
Program Calendar
Browse By Day
Browse By Room
- Information Menu
Search Tips
Virtual Exhibit Hall
- Navigation and Settings Menu
Personal Schedule
Sign In
Deepnudes among Belgian youth: how AI-apps do have it all.
Fri, September 13, 3:30 to 4:45pm, Faculty of Law, University of Bucharest, Floor: 1st floor, Amphitheater 5 „Constantin Stătescu”Abstract
Artificial Intelligence (AI) can be used for beneficial technological purposes but it also fuels several types of abuse and inflicts harm. AI has access to your personal information circulating on the Internet e.g., your personal photos. These can be easily manipulated by various AI techniques and used for malicious intents. Deepnudes or the creation of a realistic-looking nude by using AI is one of these malicious ends. A survey was conducted (N=2819) among Belgian youth (15- to 25-years). In this study, 12.8% indicated to know at least one tool to create deepnudes and 7.8% have ever tried to create them. How deepnude applications are built and the Terms of Service (ToS) they apply, contribute to the complexity of this phenomenon. A deepnude app market analysis (N=25) showed that the majority of apps do not warn against the potential creation of non-consensual content and even do not prohibit that kind of behaviour. Paradoxically, these deepnuding tools do not use this smart technique for controlling prompts and thus the images being created. Another anomaly hidden in apps’ ToS: the anonymity of the user is almost always guaranteed whilst the same tool produces fake nudes based on images published online of which consent was not given to use it for this purpose. When facial characteristics are visible, this will be a violation of the General Data Protection Regulation. As long as the victim is identifiable and thus the images are partially edited, the Belgian legislator believes that this behaviour falls within the ambit of voyeurism. Full deepnudes - where the victim is not identifiable, is not covered by Belgian Criminal law, at least when the victim represented is an adult and the deepnude is not distributed. The outcomes of this paper will contribute to insights to enhance prevention initiatives, policy, and future research.