England’s universities are being urged to rethink how they assess students as generative AI makes it increasingly difficult to establish whether submitted coursework reflects a student’s own knowledge and skills. The Office for Students, England’s higher-education regulator, has warned that institutions which do not maintain valid and reliable assessment could face regulatory consequences.
The intervention places a higher-stakes emphasis on a problem universities have been working through since widely available generative AI tools began changing how essays, dissertations and take-home examinations can be produced. The regulator said the credibility of English degrees was at risk if institutions could not show that their assessments genuinely measured student learning.
A regulatory warning on assessment
According to the Office for Students, students have reported peers using AI without detection by academic staff. The regulator’s concern is not simply that AI may be used in coursework; it is that assessment systems may no longer be capable of distinguishing a student’s own work from work substantially generated or altered by a tool.
Josh Fleming, the regulator’s director of strategy and delivery, described AI use in higher education as a generational risk to the current system. The warning signals that assessment design is becoming a central compliance issue, rather than a matter left solely to individual lecturers or departments.
The Office for Students said institutions must design measures that are valid and reliable for testing knowledge and skills. It warned that a failure to do so in response to AI could breach conditions of registration and expose an institution to sanctions.
Why conventional coursework is under pressure
Long-form written assignments have long been a major part of university assessment because they can test research, argument, analysis and communication. Generative AI complicates that model by making it easier to create fluent text quickly, including material that may appear tailored to a specific assignment.
The difficulty is not confined to essays. The regulator’s investigation also pointed to dissertations and open-book examinations as areas where institutions must consider how AI affects the evidence they receive about a student’s capability. In that environment, a polished final document alone may be less persuasive as proof of independent learning than it once was.
This does not mean every written assignment will disappear. Instead, the policy direction outlined by the regulator suggests universities will need to make deliberate choices about which tasks can still provide dependable evidence, what safeguards are appropriate, and when assessment must take place in settings where tool use can be limited or observed.
A return to more direct evidence
One response highlighted by the regulator is greater use of AI-free assessment. This can include in-person written examinations and oral examinations, often known as vivas. Such formats can give an assessor more direct evidence of how a student explains a concept, responds to questions or works through a problem.
For students, this could mean a more visible shift in the balance between coursework completed over weeks and assessment completed under supervised conditions. For staff, it may mean redesigning modules, preparing different forms of assessment and creating marking approaches that can assess both final answers and a student’s reasoning.
In-person assessment is not the only option. The Office for Students noted examples in which students are asked to critique or edit AI-generated output. That approach treats AI as part of the subject of assessment: the student must identify limitations, make informed changes and demonstrate judgment rather than simply submitting generated text.
Another model is process-based assessment. Students can be asked to document how they researched and produced their work, giving educators a clearer view of the development behind a final submission. The exact value of such evidence will depend on the course and task, but the underlying aim is consistent: assessments should reveal learning, not merely collect an end product.
Pressure for universities to act
The Department for Education said it expected universities to take the issue seriously and expected the Office for Students to enforce existing standards effectively. It also called for the regulatory framework around AI to be explicit, strong and durable.
The regulator acknowledged that many institutions have already begun redesigning assessments. That matters because universities are not starting from a single common baseline. Courses differ widely in what they teach and in how students need to demonstrate competence. A laboratory discipline, an arts programme and a professional course may each need a different mix of supervised, oral, practical and written work.
The challenge is therefore broader than detecting AI-generated writing. Detection tools can be uncertain, while assessment policy must remain fair to students and workable for teaching staff. The regulator’s emphasis on valid and reliable measures shifts attention toward designing assessments that can stand on their own, rather than relying on after-the-fact suspicion about a submission.
What changes next
The Office for Students’ position is likely to accelerate experimentation with assessment formats across England’s higher-education sector. More oral presentations, supervised writing, vivas and documented project work could become familiar features of programmes where independent authorship is essential.
At the same time, the examples cited by the regulator suggest universities are not being asked to ignore AI altogether. Asking students to evaluate or improve AI output may allow institutions to assess critical thinking and subject knowledge while recognising that the technology is now part of the wider learning environment.
The immediate test for universities will be whether they can make those changes without reducing assessment to a narrower set of easily supervised tasks. The regulator’s message is clear: institutions must be able to demonstrate that a degree represents the student’s knowledge and skills. In the AI era, that may require a more varied and more direct account of how learning is assessed.
This article is based on reporting by The Guardian. Read the original article.
Originally published on theguardian.com
