The Colonized Cannot Consent: A Decolonial Analysis of Data Rights, Data Sovereignty, and the Doctrines Underpinning Large-Scale AI Systems

Situated at the intersection of colonial history and sociotechnical analyses of AI, COLONAILISM investigates the apparent degradation of the doctrine of (digital) consent, specifically as it pertains to data collection by AI labs for the development of large-scale AI systems. Its analysis excavates the strong and instructive analogy with the period in colonial history during which the very same concept was subject to scrutiny because critics came to argue that Indigenous actors lacked the capacity to give verifiable consent in treaties involving the cession of sovereign rights. Such a revision resulted in conditions that caused 75% of the world's people to become subject to colonial powers. The analogy suggests the same power concentration/alienation has the potential to recur (if it hasn’t already) following rather similar dynamics in the scramble for user data that is accompanying the development of new foundation models. Importantly, COLONAILISM makes interventions not only in the academic discourse, but also in policy formation (at the municipal and global levels), and industry by developing new decolonial ethics, safety, and governance toolkits AI labs can use to identify and mitigate practices that are harmful to human systems or exploit their value unfairly.