Ofqual's approach to regulating the use of artificial intelligence in the qualifications sector
- Document
- 24 April 2024
- Event
- 24 April 2024
- Retrieved
- 16 September 2026
The classroom note
Since 24 April 2024, England's exams regulator has had a public position on the technology reshaping coursework and marking: Ofqual's approach to regulating the use of artificial intelligence in the qualifications sector, a policy paper most recently updated 16 July 2026, as retrieved on 16 September 2026. For a head of centre wondering who sets the rule on AI in an exam hall, the answer is layered: Ofqual states aims and boundaries for the sector, while day-to-day requirements on students and teachers are written by the awarding bodies it regulates, working jointly through bodies such as the Joint Council for Qualifications.
What the evidence says
Ofqual's own paper states five aims shaping its approach: 'ensuring fairness for students,' 'maintaining validity of qualifications,' 'protecting security,' 'maintaining public confidence,' and 'enabling innovation.' It frames AI as something that can 'support the design, development and delivery of assessments' when used responsibly. What the published material available on the page does not itself set out is a specific numeric threshold, a required disclosure format, or a named detection technology; those operational details sit instead in the joint guidance the exam boards publish, which does specify referencing formats and named checks such as comparing a submission against a student's earlier work.
The implementation question
The division of labour has a practical cost: a regulator's five aims are not, by themselves, something a teacher can apply on a Tuesday afternoon to a suspicious paragraph. Translating 'fairness' and 'security' into an actual classroom check is delegated to the boards, and from the boards partly onward again to individual teachers doing the authentication the JCQ guidance describes. That layered structure means a change in Ofqual's stated aims does not automatically change a classroom's checklist; the boards have to update their own guidance first, and centres have to notice that it changed.
What holds and what fails
What holds is a coherent division of roles: a regulator setting outcomes it wants preserved, and boards writing the operational rules that pursue those outcomes, which is how Ofqual regulates most other aspects of exams too. What is likely to fail, editorially, is an assumption that the existence of a regulatory approach document means detailed protection is already in place at classroom level; the aims are stated at a level of generality that leaves considerable room for boards to differ from each other in how they implement them.
- Which board's guidance actually governs this centre's qualifications, and has this centre read the current version?
- Does 'maintaining validity' mean the same specific thing to this board as it does in Ofqual's five aims?
- If Ofqual's approach paper is updated again, who at this centre is responsible for noticing?
A regulator's aims and a board's rulebook are different documents serving different purposes, and mistaking one for the other is an easy way to assume a protection exists that has not yet been written down for this specific qualification.
Sources & reading trail
Ofqual's five stated aims for regulating AI in qualifications, and its framing of AI's potential benefits and risks.
Source published: 24 April 2024 · Retrieved: 16 September 2026
Shows where operational detail (referencing formats, authentication checks) actually sits, distinct from Ofqual's aims.
Source published: 30 April 2025 · Retrieved: 16 September 2026
Departments, studies and vendor documents establish the record; the implementation reading and the boundary are School AI Atlas editorial analysis. This retrospective draft does not imply the site published on the event date.