In this project, Anirban Ray, Addie Sayers, and I examine the concept of archival data sovereignty with respect to generative AI. More specifically, we use qualitative content analysis to analyze the discourses of data sovereignty within three archives representing cultural, commercial, and community archival contexts: the Digital Archive of Indigenous Language Persistence, LexisNexis, and Reddit. We demonstrate how content management, user engagement, and data control vary across these archives. Ultimately, we argue that the archives’ data sovereignty with regard to AI integration varies and reveals differences of ethics and usage encoded not only in the language of policies but also in how the corresponding concepts are defined through practice. We conclude that archival data sovereignty is a rich construct that emerges from context with implications for technical communication practice.