Signals/11 Nov 2026
England regulator says AI undermines university coursework
The Office for Students says AI makes coursework hard to assess and warns universities that weak assessment could breach registration conditions and bring sanctions.
In plain words
England's higher education regulator says AI tools have made it very hard for universities to tell whether students did their own coursework. It wants more exams and in-person tests where AI cannot be used. It also warns that universities that fail to fix this could face sanctions. For students and employers, the worry is that a degree may no longer show what the holder knows. Readers judging a graduate's skills may want to look at how that person was assessed, not only at the grade.
What happened
The Office for Students (OfS), England's higher education regulator, published an investigation with Advance HE, a charity promoting university governance and teaching. The Guardian reports that the researchers interviewed more than 300 students and staff at 127 universities and colleges this year to gauge AI policies and use in teaching and assessment.
The OfS says the credibility of degrees from English universities is under threat from heavy AI use in essays, dissertations and "open book" exams. It warns that failing to design valid and reliable measures of students' knowledge and skills could breach conditions of registration and lead to sanctions. Josh Fleming, the OfS director of strategy and delivery, called AI use a "generational risk" to the system and called for "far greater use of AI-free assessment".
The OfS says many institutions have already changed assessment: more in-person written exams, oral examinations known as vivas, tasks where students critique or edit AI output, and "process-based" assessments where students detail how they researched and produced their work. A Department for Education spokesperson said the department expects the OfS to enforce existing standards effectively and to make its regulatory framework explicit, strong and durable on AI.
What is confirmed and what is not
Confirmed by the report as described by The Guardian: the size of the survey, the OfS warning about registration conditions, the regulator's call for AI-free assessment, and the Department for Education's statement.
Claimed by interviewees: that coursework is now "overwhelmingly" written with AI help, that students complete assessments in minutes and get similar grades, and that peers go undetected. These are reported accounts from staff and students, and the Guardian text gives no measured share of submissions that used AI.
Unknown: which institutions, if any, face action; what sanction would apply; what deadline applies; whether the OfS will issue new rules; and how the interviewed sample was chosen. The Guardian's report contains no response from universities or student bodies.
- 1Students at English universities and colleges8
- 2University staff designing and marking assessments9
- 3University leaders facing possible regulatory sanctions7
- 4Employers relying on degree grades5
- 5Government and the regulator setting AI rules4
Analysis
Degree credibility and graduates' trust in grades
The most direct consequence is for what a degree certifies. If coursework grades no longer separate students who understood the material from those who did not, the grade loses meaning for the student, the university and any employer relying on it. The report describes students who completed work using AI in a few minutes and received similar grades to those who did not, and peers unable to recall basic details of work they had just submitted. If that pattern is widespread, the harm falls first on honest students, whose effort is not distinguished, and then on employers who cannot read the signal. Confidence is medium: the evidence is interview-based, and the sources give no figure for how common it is.
Law and policy
The regulatory lever is the conditions of registration. The OfS says that failing to design valid and reliable measures of knowledge and skills could breach them and bring sanctions. Mechanistically, a regulator would have to show that a particular institution's assessment does not reliably measure what it claims to. That is hard when the institution can point to redesigned assessment. The sources do not state the sanctions or the standard of proof. The Department for Education statement adds political pressure and asks for a framework that is "explicit", which suggests the current rules may be less specific than government wants. That is TOSWO Research's reading of the statement, not something the sources say outright.
The parties differ in role. The OfS warns and urges. The department tells the regulator to enforce and to tighten rules. Universities, in the interviews, partly agree that some assessments are unfit for purpose. No source shows a university disputing the finding.
Technology and industry
The report notes that detection is not the main answer; the remedy it points to is changing the assessment format. Vivas, in-person exams and process-based tasks work by moving evidence of skill to a setting where AI help can be observed or excluded. Process-based assessment still depends on honest self-reporting, so it may be easier to game than a supervised exam. The sources do not test that. Reasoning from general principle, supervised formats cost more staff time and scale poorly to large cohorts, which could pressure budgets. No source gives costs.
A second effect is on the AI tools themselves. The OfS also describes using AI critique tasks as a design, so it does not call for banning AI from study. The stated aim is to help universities "reap the benefits of AI while avoiding the risks".
Daily life
For students, a return to exams changes workload and stress and may disadvantage those who do better in coursework. The sources do not discuss accessibility or fairness for students who need adjustments, a gap worth watching. For employers and the public, the practical effect is uncertainty about what a recent graduate can do unaided.
The strongest reason this view could be wrong
The central evidence is what 300-plus people said in interviews. Staff who feel overwhelmed may overstate the problem, and students may boast about undetected AI use without it being true. Also, the OfS finding that many institutions have already redesigned assessment suggests the sector may be adapting faster than the headline suggests. Evidence that would change this reading includes measured data on AI use and on grade differences, or grade outcomes that stay stable after assessment changes.
Link to wider public-sector AI oversight
The story fits a pattern in which regulators respond to AI after practice has already changed. TOSWO has covered a similar governance problem in Australia's report warning of another robodebt, though that concerns automated government decisions, not education. Readers tracking such developments can use the live AI risk monitor and the incident record.
What would settle it
The full OfS and Advance HE report would show the method, the sample and any numbers beyond the quotes. A formal OfS notice or consultation on its conditions of registration would show whether sanctions are real. Any enforcement action against a named institution, and universities' published assessment changes, would show whether the warning has force. Responses from universities' representative bodies and from student groups would test the findings.
What can be done
Universities can add supervised exams, vivas and process-based assessment, as the OfS describes. Students can keep drafts and notes that show how they produced their work. Employers can ask for evidence of skills in interviews or practical tests instead of relying only on grades. The OfS can set out clearer rules, as the Department for Education asks. For wider preparation against AI-related disruption, see the Day Zero plan.
Questions readers ask
Is the Office for Students banning AI in universities?
No. The Guardian reports that the OfS wants far greater use of AI-free assessment but also says it will help universities gain the benefits of AI while avoiding the risks. Some suggested formats even involve students critiquing or editing AI output.
Can universities be punished over AI and coursework?
The OfS warns that failing to design valid and reliable measures of knowledge and skills could breach conditions of registration and lead to sanctions. The sources do not say what the sanctions are or whether any institution is under investigation.
How many students use AI to cheat?
The sources give no figure. The report draws on interviews with more than 300 students and staff at 127 universities and colleges, and some described AI use as rife. That is testimony, not a measured rate.
What is a viva?
The OfS describes vivas as oral examinations. It lists them with in-person written exams as ways institutions have redesigned assessment to reduce AI misuse.
Written by Toswo Research. This is not investment, legal or medical advice. To report an error, see corrections.
Toswo Signals, "England regulator says AI undermines university coursework", https://www.toswo.com/signals/england-regulator-says-ai-undermines-university-coursework, published 2026-11-11, retrieved 2026-11-22.