by Leila Ballabio
As students’ summer break slowly came to an end, the beginning of the semester was afflicted with concerns on Turnitin’s changing End-User License Agreement (EULA). Many university students collectively voiced their concerns and urged their respective universities to review the new agreements and more generally, inspect and critically evaluate the use and role of AI-detection software in higher education.
The enforcement of the new EULA policy was set to begin in July 2026. The proposed updates would allow student submitted work to be used to train Turnitin AI models, which provoked a major uproar. Student unions and universities consequently addressed concerns relating to intellectual property rights and a lack of meaningful consent. In the United Kingdom, approximately 98% of universities use Turnitin products. Notably, Universities such as Southampton, Newcastle, and Westminster have reconsidered their relationship and terminated contractual agreements with Turnitin. This led to Turnitin issuing a statement on its new user policy and pausing the implementation temporarily: “we do not train any generative AI models, or LLMs. We only train a classifier model for AI detection in our Turnitin Originality license.”
Currently, the policy change has been postponed for September 2027, as Turnitin aims to adapt its policy in closer alignment with customer wishes; however, some of the damage has already been done. According to a competitor of Turnitin, approximately 50 universities have opted-out (at least partially) from using the software. Notably, most changes have occurred in the modern Anglo-Saxon sphere, with an initial flood of cancellations happening in 2023 – due to the automatic implementation of an AI-detection tool – and in 2026.
So where does the Geneva Graduate Institute fit into this dynamic?
Turnitin was implemented at the Institute originally as a plagiarism detection software and it fulfilled that role quite well. In 2023, as the AI-detection tool was added, the university’s experience with the software was unsatisfactory. According to a source within the Institute, the tool is conditional and unreliable. Put simply, AI-detection is simply quite faulty. There are many studies available to corroborate this perspective and there is an increasing difficulty for the Institute to adjust to the rapid growth of AI and to evaluate the effectiveness of Turnitin.
Generally, there has been dissent with the implementation of Turnitin at the Institute: the validity of the AI-detection software has been scrutinised due to a lack of metrics and examples of indicators of AI in a submission. As a result, the Institute is looking into finding an alternative software. Furthermore, the question of students’ intellectual property rights has been raised and will be addressed within this quest for an alternative tool. This importance of intellectual property rights emphasises the need for meaningful consent and how students deserve to have agency over their own work.
There is an increasing worry that AI models are being trained with data belonging and generated by individuals. In an article by the University of Sheffield, student responses show discomfort with the increased influence of AI on their intellectual property. To quote a student of theirs, “AI was never a choice, it just appeared, replaced everything and demanded us to be okay with it.”
Furthermore, echoing some of the concern of my fellow students at the Institute: “I am just tired of always having to talk about AI,”
There seems to be some type of AI fatigue plaguing the academic sphere. Research has shown that false positives and the focus on AI in academia causes significant stress and anxiety amongst the student body: whilst Turnitin boasts a false-positive rate of under 1%, this results in over 100’000 assessments being incorrectly flagged in the Australian higher education sector alone. Turnitin asserts the position that its software must not be used as an isolated determinant. To which extent a software can accurately verify the contents of generated material is up in arms; however, it also opens doors to alternative solutions to supervise and assess student work.
To follow up on this, a roundtable was held last week on the following topic: Power, Pens and Prompts: How AI Is Reshaping Knowledge Production in International Relations and Political Science. The shared perspectives of Dr. Christiana Parreira, Assistant Professor at the Geneva Graduate Institute, Dr. Wallace Cheng, Professor and Programme Director for Frontier Technologies and Governance at the Geneva School of Diplomacy, and Ines Gassal-Bosch, Senior Research Officer in the Geopolitics and Global Futures Department at the Geneva Centre for Security Policy, concluded that the future of academia and knowledge production requires creativity. Universities and individuals must adapt to the changes in technologies, and find more creative solutions to assessing individual work, and balancing the tools of artificial intelligence. Professors themselves are grappling on how to adjust their course curriculum to more adequately evaluate their students’ work, but also use the insights of their academic fellows to judge their own implementation of AI.
In an effort to highlight students’ voices, The Graduate Press is eager to hear from fellow students: what concerns do you have about AI-detection software at the Institute? What changes do you wish to see made? We invite you to comment your thoughts below.

0 comments on “AI Fatigue? Turnitin and AI-Detection Softwares Under Fire”