Media archives and the algorithms trained on their data do not merely document reality but they also enable the categorisation, surveillance, and control of people and groups. Images, recordings, and metadata actively structure how individuals are rendered visible, classifiable, and governable. Historically, colonial archives have reinforced stereotypes and power relations through racialised, gendered, and normative forms of representation. Today, as datafication intensifies, these dynamics are extended and increasingly automated. This workshop explores how subject formation is shaped alongside evolving media technologies. From photography and broadcast television to infrastructures like CCTV, drones, facial recognition, and AI systems, technologies capture, classify, and circulate images of people in specific ways. These systems rely on datafication, translating individuals into data stored in datasets that enable automated sorting and prediction. As AI models develop, technology companies are turning to specialised media archives as potential training data sources. Yet archives are not neutral: they reflect historical power dynamics, institutional priorities, and longstanding patterns of representation. When such materials enter machine learning pipelines, existing stereotypes, omissions, and biases may be reproduced or amplified, shaping how people are recognised and treated.