The advancement of Human-Centered AI (HCAI) is dependent on access to large and diverse datasets. However, the mechanisms for acquiring this data are plagued by a fundamental crisis of trust and a lack of respect for data sovereignty, creating a vicious cycle that starves AI systems. This paper introduces the Sovereign Loop, a novel model that resolves this tension by architecting systems where trust is not a prerequisite but a dynamically generated outcome of human-centric design. We argue that embedding sovereignty-by-design is the critical first step in a virtuous cycle: it enables trust calibration, which facilitates the flow of high-quality, consented data. This data fuels the development of robust HCAI systems that are, by architectural imperative, designed to amplify human agency through explainable, collaborative interfaces. The positive experience of this human-AI partnership feedback is to strengthen trust, justifying further data sovereignty. We formalize this feedback loop and conceptually validate its hypotheses through an agent-based simulation. Our work illustrates how sovereignty can be the human dimension of HCAI by embedding agency and consent directly into system design.