Focusing on the technical core of how AI learns, this chapter details how data serves as both a mirror and amplifier of social inequities. It classifies common types of bias—selection, representation, and measurement—and shows how these distortions arise from flawed archives, incomplete datasets, and cultural blind spots. The chapter explores data curation strategies (proactive auditing, balanced sampling) and post hoc techniques (algorithmic reweighting, fairness constraints), underscoring that addressing bias is as much an ethical and philosophical challenge as a technical one.