The Pangram Backlash Unfolding on College Campuses

Four years into the ChatGPT era, colleges still have no settled way to handle AI-written student work. Professors have tried detectors and tricks such as embedding hidden instructions in assignments, but students adapt, and the consequences of mistakes can be serious. Turnitin, used by roughly 1,400 North American institutions, says its detector finds some AI writing in nearly half of university submissions, while acknowledging that it misses some AI work and can falsely flag human writing. Newer tool Pangram claims far fewer false positives and has performed well in independent testing, but it has not yet become standard on campuses, where policies often reflect early research on less accurate detectors.

Even improved systems are not dependable proof: researchers found that “humanizer” software can disguise AI text, and detector results may be affected by how students edit their work. False accusations can damage students’ prospects, raise privacy concerns, and disproportionately threaten people using language-editing tools; institutions and vendors therefore caution against treating a score as conclusive evidence. Meanwhile, some educators argue that students should learn to use AI responsibly rather than be policed, while others worry that outsourcing assignments weakens learning. MIT’s recent report calls for a broader redesign of education to encourage reasoning and meaningful work, not merely completion for grades. Such changes demand time and resources many instructors lack. AI cheating is most acute in large introductory courses with routine assignments, while advanced courses built around original research are less vulnerable.