New research from the Office for Students (OfS) and Advance HE found that many institutions still use unsupervised assessments that can be completed to a high standard with AI tools "without substantive engagement from students." In roundtables and interviews, students described peers who finished assessments with large language models, "sometimes in just a few minutes," and got grades similar to students who did the work themselves.

The regulator says closed-book exams, presentations, practical assessments and performances are, in many cases, likely to give the most valid measure of what students know. The research also found that many students struggle to understand their institution's AI policies, and that AI literacy training for staff is often voluntary, with low participation.

"It also presents a generational risk to the system as it stands, threatening to undermine students' learning and institutions' ability to assess them," said Josh Fleming, the OfS's director of strategy and delivery. "We need to see far greater efforts to tackle these risks, including far greater use of AI-free assessment."

The U.S. parallel: American math departments are already moving grades out of homework and into the room. See our case study on what math teachers grade now.