Universities are increasing use of oral exams and other in-person defenses as generative AI makes take-home work easier to replicate while weakening students’ ability to explain their reasoning. Cornell faculty members are piloting oral defenses designed to verify learning rather than prevent cheating, and the trend is expanding at other institutions. The University of Pennsylvania is pairing oral exams with written papers, according to faculty and teaching center leadership, citing concerns that students are losing cognitive capacity and creativity when assignments can be outsourced to AI. Across campuses, faculty workshops and assessment redesign efforts focus on how to measure genuine understanding when standard written submissions are less reliable. The reported shift reflects a larger assessment governance challenge: institutions must balance academic integrity requirements, equity, and workload while ensuring students can demonstrate what they know.