New analysis argues that AI-driven teaching support may undermine learning when instructors delegate core instructional work to AI systems. The report cites a randomized trial described as “Generative AI Can Harm Teaching,” in which teachers with access to a ChatGPT-based teaching assistant customized to Turkey’s national curriculum saw less motivated student responses and lower standardized final-exam scores. Higher education and K-12 leaders are likely to treat the finding as a caution for instructional design—particularly how AI is deployed for lecture notes, assignments, exams, and feedback generation. The report adds that effects were more pronounced for students taught by weaker instructors. The development increases urgency for teacher training on responsible AI use, emphasizing instructional oversight rather than AI replacement.