Researchers and vendors are reporting linked shifts in how students write and how faculty teach. A multi‑institution team led by University of Washington Ph.D. student Liwei Jiang quantified “inter‑model homogeneity,” finding that responses from more than 70 large language models often converge on the same metaphors, imagery and punctuation. The study signals a new academic‑integrity challenge for instructors especially in creative and open‑ended assignments. At the same time, learning‑management provider Canvas launched an AI teaching agent intended to automate routine faculty tasks while stopping short of full grading automation. Canvas said the agent will save time on “low‑value tasks,” but professors and academic‑integrity experts warn that agentic systems may further flatten student voice and complicate plagiarism detection. Expect colleges to update academic‑honesty rules and assessment design as both model output and campus tools evolve.