At Cornell University, biomedical engineering instructor Chris Schaffer is using oral “defense” exams to make it harder for students to outsource learning to generative AI, requiring direct explanation to an instructor with no writing tools or technology. The story describes broader adoption by faculty across U.S. campuses, pairing oral evaluations with written work and emphasizing that students can no longer rely on perfect take-home outputs if they cannot explain their reasoning face-to-face. The development matters for institutional assessment policy because it changes how academic integrity is operationalized—shifting from detecting cheating toward verifying conceptual mastery under real-time questioning.