OfS tells universities to use AI-free assessment: secure tests must show who did the thinking

The Office for Students (OfS), the regulator for higher education in England, published "Making AI work for students: reflections on institutional practice" on 11 November, produced with Advance HE. The research ran from June to September 2026 and draws on conversations with over 300 students and staff from over 120 institutions. It finds that unsupervised assessments can often be completed to a high standard with AI, that AI detection tools "do not consistently and reliably identify AI use", that students are often unclear about their university's AI policy, and that AI literacy training for staff is often optional and little taken up.

The OfS says it will require assessments to be "valid and reliable measures of students' knowledge and skills", and that "in many cases, this will require the use of secure (AI-free) assessments", such as closed-book exams, presentations, practical assessments and performances. It says it will hold universities to account through conditions of registration, monetary penalties, or suspending student support or grant funding. Josh Fleming, its Director of Strategy and Delivery, says AI presents "a generational risk to the system as it stands". Advance HE will launch an AI-focused Change Academy in early 2027.

This matters well beyond universities. Every student who sat coursework this year, and every family paying for a degree, is affected by whether a grade can be trusted. Employers rely on the same degrees as evidence of what a graduate can do. Schools and exam boards face the identical question at GCSE and A level, and the regulator's answer will shape what they are asked next.

Dan Fitzpatrick's view is that the regulator is right about the problem and only half right about the cure. Secure assessment protects the value of a grade, and it is a sensible short-term step. But detection has failed, which the report itself says, and a room full of people with pens cannot be the whole plan. A graduate who has never learned to work well with AI is as poorly prepared as one who has outsourced all their thinking to it. The task is to design assessment that shows how a person thinks, with and without the tools, and to teach AI literacy so that students and staff know where the line sits.

What leaders should do now. School and college leaders should expect the same questions at 16 and 18, and should review which tasks genuinely show a pupil's own thinking. Teachers can add short live elements to coursework, such as a conversation about how the work was made, rather than relying on detection software. Parents should ask their child's school or university in plain terms what AI is allowed, where, and why. Employers should ask graduates to explain their reasoning as well as show their output, and should say so in recruitment.

Dan Fitzpatrick works with schools, educators and leaders on adopting AI with judgement, on assessment and AI literacy, and on keeping people in charge of the thinking, and he writes and speaks regularly on the questions this report raises for learning, work and trust.

Dan Fitzpatrick is available for interview and comment on this story. He is available today for live or pre-recorded radio, TV, podcast and print interviews, with a broadcast-quality remote setup for live or pre-recorded down-the-line interviews. Contact: dan@theaieducator.io