Unequal Inputs, Unequal Outcomes: The Human Rights Risks of Generative AI