The Office for Students (OfS), which oversees higher education in England, has told universities that they must adopt assessment methods that cannot be gamed by artificial intelligence. The regulator says it may impose sanctions on institutions that do not move away from coursework that can be produced by tools like ChatGPT. The warning follows a joint OfS‑Advance HE survey of more than 300 respondents across 127 colleges. The survey found that many students rely heavily on AI for essays and dissertations, often without being caught, and still earn grades comparable to peers who do the work themselves.
Students reportedly submitted whole papers generated by ChatGPT and then struggled to recall basic details from the text. Lecturers described seeing large batches of assignments that shared identical sources, phrasing, and style – a pattern that suggests machine‑assisted writing. OfS strategy director Josh Fleming called the trend a “generational risk” that could erode confidence in university degrees.
In response, the regulator is urging a return to traditional, in‑person assessments such as timed exams, oral presentations, and tasks that require students to critique or edit AI‑generated drafts. The Department for Education has said it expects the OfS to make its regulatory framework on AI use “explicit, strong and durable.” Several universities have already begun piloting oral exams or project‑based assessments that are harder for AI to complete on its own.
What this means for students and staff
If the OfS follows through with sanctions, universities may face reduced funding or other penalties until they can demonstrate reliable AI‑free assessment practices. For students, this could mean fewer opportunities to submit essays that rely on generative tools and a greater emphasis on live, interactive testing. Staff will likely need to redesign courses and develop new marking rubrics that focus on critical thinking and originality rather than length‑based essays.
Why it matters
Students could see a shift toward more oral exams and live presentations, reducing the chance to hand in AI‑written work. Universities may need to overhaul assessment design, and failure to do so could trigger funding penalties from the regulator.
