This study shifts the AI debate from institutional governance to lived epistemic position. It compares students from the Global North and Global South and asks how generative AI mediates access to knowledge, authority and epistemic justice in higher education. The strongest insight is asymmetry. For some students, generative AI is a convenient auxiliary; for others, it becomes a necessary bridge across paywalls, infrastructural scarcity, limited institutional access and academic inequality. The same tool therefore appears differently depending on location, resource distribution and historical position. Convenience in one place can be compulsion in another. The paper is careful: generative systems can democratise access while also acting as digital colonisers. They may offer summaries, translations and research pathways, yet they can also reproduce dominant epistemologies, flatten context, perform authority without grounds, and obscure the limits of their claims. This double movement is crucial. The issue is not whether AI is good or bad for education in the abstract but whether it increases epistemic agency or deepens dependency. Students from the Global South may gain access to otherwise unavailable material, but the price may be reliance on systems whose knowledge hierarchy, source selection and cultural authority remain opaque. The paper therefore sharpens the concept of epistemic accountability: a system should not merely answer but disclose limits, qualify claims and resist false authority.