Safety anchors and deflected desires: generative AI and the production of sexualities

Abstract

Introduction. Many images generated by AI appear sexual in nature, and a great deal of attention is paid in computer science to preventing systems from producing some forms of sexual content. Yet, despite the fact that sex seems to be everywhere in generative AI, we are saying little about how AI might intervene in our sexualities.

Approach. This exploratory essay argues for the need to consider sexuality in relation to generative AI and draws on broad arguments contained in Michel Foucault’s work on sexuality to set the stage for future work. Drawing on Rieder’s concept of algorithmic techniques, it considers current methods of prohibiting the generation of sexual content and uses these as a base for discussing the ways that AI might play a role in the formation of sexualities.

Conclusion. This essay aims to provoke future work by sketching out some aspects of generative AI that might be relevant to studies of sexuality. Primarily, it highlights the processes of evasion and denial as potentially drawing attention to and concretizing material or concepts that are considered taboo. These observations are not intended to be validated or generalizable; they are instead intended only to illustrate the potential of future work.