The rise of generative artificial intelligence raises questions about basic assumptions concerning how learning is understood and measured. Large Language Models now perform many of the complex cognitive tasks that traditional taxonomies use as markers of higher-order human thinking. This development diminishes the usefulness of Bloom’s taxonomy, since the artifacts it treats as evidence of learning can now be produced without significant cognitive involvement. The inadequacy of Bloom’s hierarchy for characterizing learning in the postdigital era calls for new approaches. In response, we develop the Epistemic Sovereignty Model (ESM): a theoretical framework that reconceptualizes the pedagogical relationship between human learners and AI systems. Drawing on Actor-Network Theory, the ESM extends the classical didactic triangle into a tetrahedron comprising the learner, teacher, content, and AI, and traces four force flows—generative, personalized, mediatory, and extractive—through which agency and effort circulate in AI-mediated classrooms. In contrast to instrumental approaches that view AI as a tool, the ESM conceptualizes AI as a pedagogical actant that fundamentally lowers the cost of creation, redistributing cognitive labor across the network. We articulate this redistribution through the Inverse Effort Principle: as the computational cost of generation approaches zero, the cognitive investment required for meaningful validation must increase. Building on this principle, the ESM reorganizes the taxonomy around six levels, replacing Creation at Level 5 with Melioration—the strategic improvement of machine-generated baselines—and elevating epistemic vigilance and ethical reasoning to the new apex of learning. The model has direct implications for assessment, suggesting a shift from product evaluation toward process auditing and synchronous verification. The ESM offers both a theoretical framework and a practical vocabulary for cultivating subjectification in the postdigital world.