A Brown University professor responding to an unusually high-scoring midterm said he suspected AI cheating after results clustered tightly around near-perfect scores. Roberto Serrano wrote that when he ran the exam questions through a large language model, ChatGPT produced an odd, overly specific solution path similar to what multiple students used. Serrano’s response also reflects a shift in how instructors are adapting assessment formats: he adopted take-home exams for the first time following a deadly December campus shooting. He now argues those assignments cannot be trusted to measure learning when AI access and generation tools are readily available. The episode highlights the enforcement challenge for higher education assessment integrity—particularly when assessment changes intended to address safety concerns collide with the reality that AI can generate plausible answers at scale. For universities reviewing academic integrity policies and assessment design, the case adds urgency to AI-aware proctoring, assignment structure, and faculty guidance.