Deepfakes and Contextual Integrity Violations

This paper conducts a comparative analysis between intuitively ethical and unethical uses of deepfakes to characterize ethical boundaries on deepfake creation. We make a novel argument that a key condition on the permissible use of deepfakes involves preserving the contextual integrity of a deepfake's source materials. Questions about contextual integrity dovetail with traditional ethical concerns relating to data ownership, consent, privacy, trust, and transparency. We argue that this is because such concerns arise when contextual integrity is violated, making the preservation of contextual integrity an ethical priority in deepfake creation. We further explore the question of information in public versus private contexts through the novel application of “curtilage“—a concept traditionally applied to define the boundary around physical spaces afforded privacy protections—to address the blurred line between public and private life. We argue that digital spaces and the information found therein deserve a similar boundary, or “digital curtilage”. The notion of curtilage becomes a framework to consider how deepfakes should respect these digital boundaries. Lastly, the paper also examines current legal frameworks, exposing regulatory gaps that fail to adequately address the nuances of deepfake technology.