England’s higher education regulator is urging universities to ensure assessments measure learning reliably, as generative AI can produce high-scoring work without showing whether students have the skills being assessed.
A study published Nov. 11 by the Office for Students (OfS) and Advance HE drew on discussions with more than 300 students and staff across over 120 institutions from June to September 2026. Interviewees recounted classmates using large language models to finish assessed work in minutes and earning marks comparable to peers who worked unaided. The report also says many students are unsure what their institutions allow, while current AI detectors are unreliable ways to determine whether a student used the tools.
The report describes both potential benefits and risks. AI may support neurodivergent students, students with disabilities and those whose first language is not English. But using it to replace reading, research or problem-solving can reduce the effort involved in learning, the report says.
For many learning goals, the study says, supervised formats such as closed-book exams, presentations, practical assessments and performances may offer more dependable evidence when designed to exclude AI. It also supports assignments that permit AI use, provided they still test the required learning outcomes.
Josh Fleming, the OfS’s director of strategy and delivery, called for “far greater use of AI-free assessment.” The regulator said it may intervene when evidence indicates a provider is not meeting its requirements. Possible measures include registration conditions, monetary penalties and suspending access to student support funding or OfS grant funding.
The OfS plans further research into AI’s effects on higher education. Advance HE plans to launch its AI-focused Change Academy program, a peer-learning and action-learning initiative for member institutions, in early 2027.
