MIT report says AI can complete almost any undergraduate assignment it sets

MIT’s ad-hoc AI committee reports that current models can credibly complete virtually any undergraduate assignment—essays, math and science problems, proofs, and coding tasks—and have already altered campus behavior, reducing office-hour attendance, online-discussion participation, and study-group formation. In response, institutions such as Chicago Law School and Princeton have tightened exam supervision, while the EU’s AI Act classifies student-monitoring systems as high-risk and bans emotion-recognition tools in education.