Media Analysis: What and Who Can Universities Trust to Police AI Use?
Updated: Sep 7
In this blog entry, we actually review two related media reports...
Original Article 1: I was accused of using AI in my dissertation but I wrote it all myself
Publisher: BBC III Author: Marta Leshyk III Publication Date: 11th August 2026
Original Article 2: Southampton dumps Turnitin over use of students’ work to train AI
Publisher: Times Higher Education III Author: Georgia Luckhurst III Publication Date: 19th August 2026
Summary of the original articles:
Two recent UK stories highlight very different challenges facing universities as they respond to generative AI.
The first concerns Harrison Sharples, a medical student at the University of St Andrews who was accused of using AI in his dissertation despite maintaining that he had written it himself. His work was questioned by a human marker who identified features including a polished and uniform tone, consistent grammar, repetitive phrasing and a formulaic structure. The allegation was initially upheld, but following an appeal, the university ultimately cleared him and awarded his full mark – albeit with a significant delay to his graduation and obvious distress.
The second story concerns the University of Southampton's decision not to renew its Turnitin contract beyond the 2026–27 academic year. The university cited concerns about proposed changes involving the use of student submissions in developing AI services. Turnitin disputed suggestions that it uses customer or student work to train its in-house composition tool, while acknowledging that anonymised submissions may be used to improve its detection and assessment tools.
On the surface, the stories appear unrelated.
One concerns a student and the judgement of academic staff. The other concerns an institution and its relationship with a technology provider.
Yet together they reveal a common issue at the heart of the university response to generative AI: trust.
Why it matters for A4-AI (Academics for AI):
Universities have a legitimate responsibility to protect academic integrity. As generative AI makes it easier for students to produce sophisticated written work, institutions need ways of addressing inappropriate use without undermining the credibility of their assessments.
The St Andrews case illustrates the difficulty of relying on human judgement alone. The concerns about Sharples' dissertation were based on characteristics of his writing rather than evidence from an AI detection system. A polished and consistent writing style, repetitive phrasing or a formulaic structure are now, apparently, grounds to raise questions about inappropriate use of AI. Let that sink in....being good at writing, could now raise suspicions.
The initial judgement was subsequently overturned on appeal. Sharples ultimately received his full mark, but the process had already affected his academic progress and delayed his graduation. The case demonstrates how consequential it can be to not distinguish between suspicion and evidence when assessing whether AI has been used.
The Southampton story highlights a different dimension of trust.
Universities increasingly rely on third-party platforms to support assessment and academic integrity. These technologies may offer valuable tools for dealing with generative AI, but they also raise questions about how student work is handled and how those relationships with third party tech providers are governed.
Taken together, the two stories highlight three different forms of trust that universities increasingly need to consider:
Trust in judgement. Can academic staff make sufficiently reliable assessments about whether AI has been used?
Trust in technology. How much confidence should universities place in tools designed to identify or manage AI use?
Trust in providers. How should student work and data be governed when third-party companies are involved?
There is also a fourth element: trust from students. Academic integrity systems ultimately depend on students believing that allegations of inappropriate AI will be based on fair and credible evidence, and that the dealing of such allegations will be done with transparent due process and minimal delays.
This may be why the debate about AI and academic integrity needs to move beyond detection.
Rather than relying solely on attempts to identify AI use after the event, universities may need to place greater emphasis on assessment design that makes students' learning, reasoning and application of knowledge more visible. Oral examinations, presentations, supervised work, practical assignments and iterative assessment are among the approaches that could reduce the importance of determining retrospectively whether a piece of work was produced by AI.
AI may still have a role in supporting academic integrity. But technology alone cannot resolve the underlying issue.
The challenge for universities is to build an academic integrity system that is credible to institutions, defensible to academics and trusted by students.
As generative AI becomes increasingly embedded in higher education, trust may become just as important as detection.
Questions for Academia:
How reliable should evidence of AI use be before it can contribute to an academic misconduct case?
Should universities rely on AI detection or human judgement to police AI use?
What are the acceptable terms of use by third party assessment platforms which process student work?
Should universities focus more on redesigning assessment than detecting AI use?
Comments