UK regulator warns universities to protect degree standards as AI reshapes student assessments

Universities in England need to rethink how they assess students as artificial intelligence makes it increasingly difficult to establish whether submitted coursework reflects an individual's knowledge and abilities, the country's higher education regulator has warned.

The Office for Students (OfS), which regulates higher education in England, has called on universities and colleges to make greater use of secure assessments that can reliably demonstrate students' learning, including examinations conducted without access to AI tools.

The warning follows a joint research project by the OfS and Advance HE, which examined how artificial intelligence is changing teaching, learning and assessment in higher education. Published on November 11, the report, Making AI work for students: reflections on institutional practice, draws on conversations with more than 300 students and staff from over 120 higher education institutions.

The research found that many institutions continue to rely on unsupervised assessments that students can complete to a high standard using generative AI without necessarily demonstrating the knowledge and skills the assignments are intended to measure.

Students involved in the discussions described instances in which peers had used large language models to complete assignments in a matter of minutes and received grades comparable to those awarded to students who had completed the work themselves.

The findings have raised concerns about the reliability of conventional coursework, the fairness of assessment and the credibility of university qualifications.

AI is challenging conventional methods of assessment

The report identifies assessment as one of the areas where the rapid adoption of generative AI has created significant challenges for universities.

Take-home assignments, essays and other unsupervised tasks can now be completed with substantial assistance from AI tools. While these formats have traditionally allowed students to demonstrate research, analysis and writing skills, the technology has made it harder to determine how much of the submitted work reflects the student's own understanding.

The research also highlights the limitations of relying on AI detection software to identify unauthorised use. Such tools do not consistently and reliably establish whether a student has used AI, making them an uncertain basis for disciplinary decisions.

The OfS argues that universities should place greater emphasis on designing assessments that demonstrate what students know and can do, rather than attempting to identify AI use after work has been submitted.

In many cases, the regulator says, secure assessments that prevent or restrict access to AI are likely to provide a more reliable measure of student attainment. These could include closed-book examinations, in-person presentations, oral examinations, practical tasks and performances.

However, the report does not suggest that every assessment should return to traditional examinations. Universities can also develop approaches that incorporate AI under controlled conditions, provided the assessment still establishes whether students have achieved the intended learning outcomes.

Universities face pressure to protect academic standards

The OfS has made clear that institutions remain responsible for ensuring the quality and reliability of their assessments, even as AI becomes more widely available.

It expects universities and colleges to ensure that assessment methods provide valid and reliable evidence of students' knowledge and skills. Courses must also remain up to date and equip students with capabilities relevant to their future studies and employment.

The regulator has said it will hold institutions accountable where they fail to meet these expectations. Its existing regulatory powers include imposing specific conditions of registration, monetary penalties and suspending access to student support funding or OfS grant funding.

The announcement does not introduce a blanket prohibition on AI use or a new AI-specific penalty system. Rather, the regulator is explaining how existing quality requirements apply when institutions design and administer assessments in an environment where students have access to generative AI.

Josh Fleming, director of strategy and delivery at the OfS, described AI as both an opportunity for universities and a risk to students' learning and institutions' ability to assess them.

He called for greater efforts to address the risks, including a wider use of assessments that do not permit AI assistance.

The regulator will also continue research into the scale and nature of AI's impact on higher education and share examples of institutional practices that address the challenges.

Students need AI skills, not just restrictions

While the report highlights the risks of AI-assisted coursework, it does not advocate excluding the technology from higher education.

Students are already using AI to explain difficult concepts, summarise lengthy documents, generate practice questions and support revision. When used critically, these applications can help learners identify gaps in their understanding and engage with course material in ways that suit their needs.

The research also recognises potential benefits for neurodivergent students, students with disabilities and those whose first language is not English.

The concern arises when AI replaces the effort involved in reading, research, problem-solving and independent thinking. The report describes this as cognitive offloading, in which students delegate tasks to AI instead of developing the knowledge and skills those tasks are intended to build.

Students also told researchers that AI literacy would be important to their future employment prospects and that they wanted more opportunities to learn how to use the technology effectively.

The findings point to uneven provision across institutions. Some universities have developed programmes to improve students' AI capabilities, while others offer limited training or none at all. Staff training is also frequently voluntary, with participation often low.

Kathryn Harrison-Graves, deputy chief executive at Advance HE, said students wanted greater clarity about what was expected of them, while staff confidence and institutional support varied.

She highlighted the role of university leadership in establishing a shared approach to AI, communicating clear guidance and making training and resources available to both students and staff.

Institutional policies and teaching practices need attention

The research also found that students frequently struggled to understand their university's or department's rules on AI use in assessed work.

Although many institutions have introduced policies, expectations are not always communicated consistently. Students may be uncertain about when AI assistance is permitted, how it should be acknowledged and which uses could breach academic rules.

The report also identifies concerns about the use of AI by teaching staff. Some students felt that AI-generated teaching materials and feedback had affected the quality of their learning experience. Staff, meanwhile, differed considerably in their willingness and ability to incorporate the technology into teaching.

At an institutional level, AI adoption has often been driven by individual academics experimenting with tools rather than by a coordinated strategy. Some universities have established governance groups, appointed AI leads or introduced programmes to support staff, but the research found that strategic approaches remain uneven.

The report suggests that institutions need clearer policies, stronger staff development and better oversight of how AI is used in teaching, assessment and student services.

Advance HE plans leadership programme for 2027

Alongside the report, Advance HE announced plans to launch an AI-focused Change Academy programme for its member institutions in early 2027.

The programme will bring together higher education leaders from different countries to examine institutional AI strategies through peer learning and collaborative work. Advance HE also plans to convene university leaders, academics and industry experts to discuss AI in teaching, learning, governance and inclusion.

For universities, the challenge is therefore not simply to prevent students from using AI in coursework. They must establish where the technology can support learning, where it risks replacing essential academic effort and how assessment can continue to demonstrate individual achievement.

The OfS and Advance HE research suggests that institutions will need to address these questions together. As AI becomes a more established part of higher education, the credibility of degrees will increasingly depend on whether universities can show that graduates have acquired the knowledge and capabilities their qualifications represent.