Artificial Intelligence ethics is often framed in universal terms, yet such framings obscure how cultural and material practices shape fundamentally different relationships with data. In many parts of the Global South, communities interact with AI not only as users but also as repairers, annotators, and mediators of fragile infrastructures, producing forms of engagement that Western discourses on AI ethics frequently miss. Based on my long-term ethnographic and design work across Bangladesh, India, Pakistan, Canada, and other contexts, I will show how AI systems extend colonial legacies by imposing Western neoliberal values, how epistemic injustices emerge when local ways of knowing are rendered unintelligible, and how community practices such as informal data repair in Dhaka or immigrant struggles over data legitimacy in Canada reveal alternative ethical concerns. I will also highlight the growing data annotation industries in countries such as Bangladesh, India, and China, where labeling labor is promoted as a path to development but raises distinct worries around exploitation, recognition, and long-term sustainability. These accounts demonstrate that AI ethics cannot be disentangled from situated practices of data and labor, and that any attempt to globalize AI ethics must bring these lived realities to the fore in order to avoid reproducing the very exclusions it claims to resist.