Academic Misconduct and GenAI

Generative AI tools reached students faster than institutions could write policy for them. Within months of ChatGPT's public release, UK universities had to decide, case by case and often under press scrutiny, whether an AI-assisted essay was cheating, a legitimate study aid, or something in between. Three years on, a reasonably consistent sector position has emerged: unacknowledged AI-generated submissions are academic misconduct, detection is unreliable enough that policy cannot depend on it, and the more durable answer is redesigning assessment rather than out-policing the tools.

Institutional Policy: Prohibition Plus Disclosure

Bournemouth University was among the first UK institutions to rule explicitly on GenAI, declaring in February 2023 that submitting an assessment where AI software had written all or part of it, without disclosure, was an academic offence under its Policy and Procedure for Academic Offences (6H plagiarism) [1]. A revised version of the policy followed in May 2023, removing the fixed 30% Turnitin similarity threshold that had previously triggered an automatic investigation and leaving the judgement to the marker instead [2]. That pattern — permit AI as a disclosed aid, treat undisclosed AI-generated content as misconduct, and keep the threshold for triggering an investigation a matter of academic judgement rather than an automated score — has become the sector default rather than an outlier.

Individual institutions vary mainly in tone. Oxford's guidance for students, for instance, frames AI tools as something to use critically and cross-check against established sources rather than something to avoid [3], while Cambridge treats submitting AI-generated content as one's own work as straightforwardly a disciplinary matter [4]. The Russell Group's own principles, discussed below, sit above these institutional variations as a shared baseline.

Why Detection Cannot Carry the Policy

AI-text detectors, including Turnitin's, have been criticised for unreliable accuracy and for disproportionately flagging text written by non-native English speakers as AI-generated [5]; several UK universities, including Bournemouth, have for this reason chosen not to switch Turnitin's AI-detection feature on at all [2]. The consequence is that a growing share of misconduct investigations depend on oral examination rather than software output: analysis by researchers at Edinburgh Napier University's Business School, based on 2,075 misconduct cases across the sector, found that most breach investigations needing to establish whether a submission was genuinely a student's own work still resolve through a viva rather than a detector score, and quantified the very real staff time and cost that each of those investigations consumes [6]. A viva scales to a caseload of dozens; it does not scale to a caseload in the hundreds, which is where several institutions now find themselves — Times Higher Education's analysis of Freedom of Information requests across the 24 Russell Group universities found suspected AI-related misconduct cases at the University of Glasgow rising from 36 in 2022/23 to 130 in 2023/24, and at the University of Sheffield from 6 to 92 over the same period [7].

That volume is itself downstream of how many students now use these tools day to day. The HEPI/Kortext Student Generative AI Survey, based on a Savanta poll of 1,041 full-time undergraduates, found that the proportion of students using any generative AI tool rose from 66% in 2024 to 92% in 2025, and that 88% had used GenAI specifically for assessed work, up from 53% the previous year [8]. Detection built for a minority-use environment does not transfer cleanly to one where the tool is closer to the default.

Underneath institutional policy sits one piece of statute and one shared sector charter. The Skills and Post-16 Education Act 2022 made it a criminal offence in England to provide or advertise a paid "contract cheating" service — completing an assignment on a student's behalf — for students at a post-16 institution or higher education provider; critically, the Act makes clear that a student who buys such a service does not thereby commit the criminal offence themselves, only the provider does [9]. It was aimed at essay mills rather than GenAI specifically, but it sets a legal backdrop against which "AI-as-a-writing-service" sits uneasily, and some commentators have argued the drafting is broad enough to reach AI providers offering equivalent completion services, though this remains untested [9].

Above individual institutions, the Quality Assurance Agency's Academic Integrity Charter sets out seven baseline principles that UK providers are expected to build their own policies on top of, and the QAA has issued specific advice — including "Reconsidering Assessment for the ChatGPT Era" — on adapting assessment design rather than relying on catching misuse after the fact [10]. In July 2023 the Russell Group's 24 vice-chancellors adopted five shared principles: helping students and staff become AI-literate, equipping staff to support appropriate student use, adapting teaching and assessment to incorporate ethical GenAI use while protecting academic rigour, ensuring equitable access to the tools, and sharing practice across institutions as the technology moves [11]. All 24 member universities subsequently updated their academic conduct policies to reflect it.

Where This Leaves Assessment Design

The practical response converging across the sector is less about stronger detection and more about assessment that does not depend on detection working. Jisc's guidance to institutions is explicit that AI policy should not be bolted on as a separate document but folded into existing assessment, data-protection and academic-integrity policy so that the response stays coherent as the technology changes [12]. In practice that has meant a shift toward staged or in-person assessment components, explicit AI-use declarations attached to submissions, and — most durably — questions that ask for judgement, synthesis, or process rather than a free-standing artefact a model can produce end to end on request. None of that eliminates misconduct. It does shrink the space in which an unacknowledged AI submission is indistinguishable from a student's own work, which is the actual target, since perfect detection was never realistically on offer.

Our Learning Loops page covers the same territory from the other direction: what changes about learning itself when a tool can produce a plausible-looking answer on demand, and why staying "in the loop" rather than just consuming AI output is the more durable skill to teach.

  • GenAI Needs in Higher Education — what institutions need to build, beyond misconduct policy, to respond to GenAI properly.
  • Learning Loops — staying in control of AI-assisted work rather than just consuming its output.
  • Scaffolding and GenAI — designing project work that uses AI assistance without hollowing out the learning.

References

  1. Bournemouth University, "Using ChatGPT for part or all your assignment is an academic offence", 8 February 2023. https://www.bournemouth.ac.uk/news/2023-02-08/using-chatgpt-part-or-all-your-assignment-academic-offence
  2. Bournemouth University, "Important changes to our policy around academic offences", 2 May 2023. https://www.bournemouth.ac.uk/news/2023-05-02/important-changes-our-policy-around-academic-offences
  3. University of Oxford, "Using AI to support your studies". https://www.ox.ac.uk/students/academic/guidance/skills/ai-study
  4. Varsity, "Cambridge records first cases of AI-related academic misconduct". https://www.varsity.co.uk/news/28936
  5. Vanderbilt University Brightspace, "Guidance on AI Detection and Why We're Disabling Turnitin's AI Detector", August 2023. https://www.vanderbilt.edu/brightspace/2023/08/16/guidance-on-ai-detection-and-why-were-disabling-turnitins-ai-detector/
  6. Higher Education Policy Institute, "The financial impact of AI on institutions through breaches of academic integrity" (Edinburgh Napier University Business School research), 1 March 2024. https://www.hepi.ac.uk/2024/03/01/the-financial-impact-of-ai-on-institutions-through-breaches-of-academic-integrity/
  7. Times Higher Education, "Student AI cheating cases soar at UK universities". https://www.timeshighereducation.com/news/student-ai-cheating-cases-soar-uk-universities
  8. Higher Education Policy Institute / Kortext, "Student Generative AI Survey 2025", HEPI Policy Note 61, February 2025. https://www.hepi.ac.uk/reports/student-generative-ai-survey-2025/
  9. Skills and Post-16 Education Act 2022, s.15 (essay mills / contract cheating offences). https://www.legislation.gov.uk/ukpga/2022/21/notes/division/9/index.htm
  10. Quality Assurance Agency for Higher Education, "Academic Integrity Charter for UK Higher Education" and generative AI advice and resources. https://www.qaa.ac.uk/sector-resources/academic-integrity
  11. Russell Group, "Principles on the use of generative AI tools in education", July 2023. https://www.russellgroup.ac.uk/policy/policy-briefings/principles-use-generative-ai-tools-education
  12. Jisc National Centre for AI, "Navigating the Future: Higher Education policies and guidance on generative AI", 31 July 2024. https://nationalcentreforai.jiscinvolve.org/wp/2024/07/31/navigating-the-future-higher-education-policies-and-guidance-on-generative-ai/