A university introduces an AI system that predicts which first-year students are likely to drop out, using their attendance, online learning activity, fee payments and background data, and alerts advisors to contact the students flagged as at risk. In the pilot, dropout fell from 17% to 14%, but the system flagged far more students than the advisors could support, was more likely to flag students from certain nationalities, and informed students about their risk status without clear consent. As an AI specialist, critically discuss the technological strengths and limitations of this system, its ethical implications (such as fairness, privacy and transparency), and its feasibility in practice. Justify your evaluation through logical reasoning and relevant real-world examples, and recommend whether the university should continue, modify, or stop using the system.