A growing number of professors are using oral exams to respond to generative AI’s spread in higher education. Cornell University biomedical engineering professor Chris Schaffer described a return to an “oral defense” format in which students speak directly to an instructor with no reliance on laptops, chatbots, or written submissions during the evaluation. At the University of Pennsylvania, associate professor Emily Hammer is pairing oral exams with written papers, citing concerns that students are losing skills, cognitive capacity, and creativity when assignments are outsourced to AI. Penn’s Center for Teaching and Learning executive director Bruce Lenthall said faculty workshops on oral exams reflect a broader shift toward in-person assessment. The developments reflect a specific instructional governance problem: verifying learning when take-home work can be generated with convincing text. Oral exams are positioned as a method to test comprehension and process, reducing the scope for AI-only production. For institutions expanding AI guidance, this trend signals that assessment design may become one of the most contested operational areas—where academic standards, academic integrity policies, and student support expectations collide.