The Pulse

England’s university regulator calls for more AI-free assessments

The Office for Students says universities will often need AI-free assessments to measure students’ knowledge and skills reliably. Research with Advance HE found that some unsupervised assessments could be completed using AI without substantive student engagement.

AI.info Team · · fact-checked · against 3 sources ↓ · Published 12 Nov 2026

Universities in England will often need assessments that prevent AI use to measure students’ knowledge reliably, the Office for Students said after research found that some unsupervised work can be completed to a high standard with little student engagement. Its November 11 statement called for more AI-free assessment alongside efforts to help students develop skills with the technology.

The research, carried out with Advance HE, reflects the experiences of over 300 higher education students and staff from over 120 institutions. The research was conducted between June and September 2026 and focused primarily on English institutions.

Pressure on assessment

Students in interviews and roundtables described peers using large language models to complete assessments, sometimes in just a few minutes, and receiving grades similar to those who did the work themselves. Staff participants questioned whether assessments consistently and reliably measured students’ knowledge and skills.

Students are frequently unclear about whether and how they may use AI in assessed work, and it is usually not possible to establish conclusively whether they have followed their institution’s policy.

Some institutions have already redesigned assessments to reduce their susceptibility to AI, including by making greater use of closed-book exams, presentations and other controlled assessments. Approaches that integrate AI into assessments can also be positive, the report says, but need careful design to preserve their validity and reliability.

Josh Fleming, the regulator’s director of strategy and delivery, said AI presents both an opportunity and a risk to students’ learning and institutions’ ability to assess them. He called for substantially greater use of AI-free assessment.

How the rules apply

Under the regulator’s condition B4, institutions must assess students effectively, ensure each assessment is valid and reliable, and design academic regulations so that awards are credible. The guidance defines secure assessments as those designed to prevent AI use, including in-person exams, presentations, oral exams, performances and practical or applied tasks.

Where AI is permitted, the guidance identifies an approach likely to comply with the condition: assessments must still require students to demonstrate the intended learning outcomes, which may include effective use of AI. It also says universities and colleges have a legal duty not to discriminate against students and may need to adjust assessment methods for disabled students.

Where evidence suggests institutions have failed to meet requirements, the regulator can impose specific registration conditions or monetary penalties, or suspend access to student support funding or its own grant funding. The regulator also plans further structured research into the scale and nature of AI’s effects on higher education.

Skills and teaching

Students use AI to explain difficult concepts, summarise lengthy documents, generate practice questions and revise, according to the report. But students and staff also raised concerns that some uses replace the cognitive effort needed for deep learning.

The report found that students receive uneven AI literacy training: some institutions have developed programmes, while many students said they had received little or no training. Some felt staff use of AI had reduced the quality of learning materials.

Students told researchers that AI literacy mattered to their future employment prospects and that they wanted more opportunities to build those skills. Staff training is frequently optional and often has low take-up, the report found.

Next steps for institutions

Participants generally suggested that new AI-specific registration conditions could be counterproductive and constrain innovation. Some instead sought clearer guidance on how existing conditions apply to AI.

Advance HE plans to launch an AI-focused Change Academy programme for member institutions early in 2027. The peer-learning and action-learning programme is intended to bring members together to develop their institutional AI strategies.