England’s higher education regulator has concluded that too much university assessment can no longer establish whether a student actually knows the material — and it is asking universities to bring exams back into the room.
The Office for Students (OfS), which regulates universities and colleges across England, published new guidance on Wednesday, 11 November 2026 UK time, alongside the findings of a four-month study into how artificial intelligence has reshaped teaching and assessment. As reported by BBC News, the regulator found “a significant proportion” of assessments run in unsupervised, uncontrolled formats where AI can be used freely, that many “can be completed to a high standard with AI, sometimes in minutes,” and that it is “frequently possible” for students to use AI “in ways that cannot be reliably detected.”
The research — conducted with Advance HE, a teaching-excellence charity, and drawing on 244 staff and 91 students across at least 127 universities — heard that peers had completed entire assessments with AI “sometimes in just a few minutes, and received a similar grade to those who had completed the assessment themselves,” according to Times Higher Education. Some students told the regulator outright that they believed their institution’s assessments were not fit for purpose. Some staff said they were receiving piles of assignments that looked the same.
The regulator’s answer: make it survivable only in person
The OfS is not banning AI. What it has done is put teeth behind a preference. Under condition B4 — the existing requirement that institutions use effective assessment methods — the regulator now spells out what “secure” assessment means: “in-person examinations, presentations, vivas (oral examinations), performance-based assessments, and assessments that require students to demonstrate their knowledge and skills through practical or applied tasks,” per the Times Higher Education readout of the guidance. Where AI is permitted, it must be governed by “realistic, meaningful and transparent academic regulations.”
The report itself is blunter still. The OfS publication, “Making AI work for students: reflections on institutional practice”, states that secure approaches “will often be the most effective forms of assessment because they provide a high degree of confidence that students are demonstrating their own knowledge and skills,” and adds a line that will be quoted in faculty meetings for years: “If it becomes possible to complete an assessment using AI without the knowledge and skills being tested, the assessment should be replaced.”
There are consequences attached. Under proposals now out for consultation, universities that fail to meet OfS quality standards could be blocked from charging increased tuition fees or have caps imposed on their student numbers. Josh Fleming, the OfS’s director of strategy and delivery, called AI “a generational risk to the system as it stands, threatening to undermine students’ learning and institutions’ ability to assess them,” and said the sector needs “far greater use of AI-free assessment.”
The sector pushes back
The response splits along a familiar seam. Universities UK, representing more than 140 institutions, said many universities had already made “substantial changes to how they assess students. This includes more in-person exams, presentations, vivas, process-based assessment and assessments that integrate AI in a controlled way.” The National Union of Students went the other way entirely: Lewis Wilson argued AI is threatening “traditional” ways of learning that students and staff “have long said is not fit for purpose,” and said it was time to retire outdated methods rather than entrench them.
The OfS’s own findings capture the same split inside classrooms. Staff described a “polarised atmosphere” between students using AI and “conscientious objectors” who refuse it on environmental or ethical grounds — students who, under unsupervised take-home assessment, are effectively graded on their willingness to do the work the hard way while the grade signal tells them it doesn’t matter.
Why this matters beyond England: remote online exams were rolled out widely during Covid, and Swansea University research cited by the BBC suggests 93 UK universities still use them in some form — the pandemic’s assessment shortcuts have quietly become permanent infrastructure, and AI only made the loophole load-bearing. The UK Pulse’s summary of the enforcement consultation notes the regulator is signalling “the era of passive oversight regarding academic integrity is ending.”
New Zealand sits one step behind on the same path. There is no national regulator issuing assessment guidance of this kind; instead, secondary teachers told RNZ that AI misuse is a “big problem,” that detection tools “could give wildly different estimates,” and that the Ministry of Education and NZQA have advice for teachers but no national guidance — exactly the policy vacuum the OfS study identified, with one in five students in one Auckland teacher’s class misusing AI for assessments as early as last year. New Zealand universities, meanwhile, have been running the same scramble toward supervised assessment as their Australian counterparts, as we covered when ANU staff called their university’s response “hysterical”.
Detection did not work, so the venue is the control
The quiet collapse underneath all of this is detection. The OfS found that “AI detection tools do not consistently and reliably identify AI use” — which closes off the industry’s favourite escape hatch. If you cannot prove a piece of work was machine-written, the only remaining integrity check is making the work happen where a machine can’t quietly do it: a timetabled room, a pen, an oral defence.
That is the same conclusion American colleges reached, as we covered when US colleges started replacing written exams with oral assessments, and it mirrors what happened when exam season became a global AI story earlier this year. England’s version adds something those earlier stories lacked: a regulator willing to tie assessment security to institutional funding and enrolment limits. The wager is that a degree’s market value is a shared asset, and that no university can be trusted to protect it alone when the cheap path — let everyone use AI, hand out the same grades — is available to every competitor.
Whether vivas and closed-book exams can scale across a mass higher-education system without crushing the student experience is the obvious objection, and the NUS is right that the pre-AI lecture-essay-slide model had its own problems. But the OfS report makes the counter-argument neatly: “many students now see AI as the most effective route to a high grade, and students may feel they are at a disadvantage if they do not use AI.” An assessment system that punishes honesty is already broken. England has decided to fix the venue before it fixes anything else.
Sources: BBC News, Office for Students — Making AI work for students, Times Higher Education, RNZ, The UK Pulse
