A randomized controlled trial in Turkey found that high school students given unrestricted access to generative AI for math practice performed worse on later independent exams than peers without AI ...