This paper explores the epistemological repercussions associated with incomplete datasets in machine learning and subsequent artificial intelligence applications. Extensive literature has emerged in the last decade showing how data-driven computing technologies, such as machine learning, tend to reproduce structural inequalities due, in part, to biased datasets used in their training. This bias and its implications have often been explored from the analysis of constituent items in data sets—for example, the predominant representation of one group over another in training and benchmark datasets. This paper explores a similar yet distinct issue in dataset curation at the center of machine learning technologies: the implications of missing data. In particular, this paper focuses on how such omission may fail to encode significant cultural values into the computing architecture of AI and considers the epistemological challenges this poses, referencing what Sabina Leonelli calls data imaginaries (2021). Through this analysis, this paper argues that missing data is not simply a technical failure or oversight but an ethical and epistemological problem to be addressed. This paper helps elucidate and articulate how examining where and why missing data leads to unique forms of cultural loss perpetuated—or at least uniquely exacerbates through scale, prowess, and speed—data-driven computing technologies, such as machine learning, as they occupy more and more social spaces.