New polling from Instructure finds AI use is already widespread in college classrooms: 90% of students report using AI at least occasionally, and 61% of instructors say they do the same. Yet only 11% of higher ed instructors report receiving comprehensive AI training, and 41% say they have not received any formal training. The survey also shows persistent concerns about reliability: 65% of both students and educators worry AI can sound confident when it is wrong, with additional anxiety about overreliance and loss of critical thinking and deep learning. For institutions, this combination—high adoption with low training—creates near-term governance pressure. Colleges are likely to expand faculty development, clarify expectations for acceptable AI use, and update academic integrity policies to match actual student behavior. Instructure’s findings suggest universities are shifting from “whether to use AI” to “how to train, supervise, and set boundaries,” with practical support becoming the differentiator rather than policy alone.