Almost every essay written at a British university passes through the same piece of software. Students upload their coursework to Turnitin, it is compared against a vast archive of previous submissions and published work, and a similarity score comes back. Students at the University of York have now raised a question about the arrangement that is harder to answer than it looks: once your essay is in that archive, what else can it be used for?
What students are worried about
The concern is about changes to the terms of the licence under which universities buy the service, and whether those terms could permit student work to be used in training artificial intelligence. Sam Dickinson of York Students' Union framed it as a question of consent: students hand over their work because they are required to, as a condition of being assessed, not because they chose a product and agreed to its terms. There is no version of submitting coursework in which a student can decline and still graduate.
Dickinson also connected it to a bleaker backdrop — graduate vacancies are falling, and students are being asked to contribute, without payment or choice, to systems that may narrow the job market they are about to enter.
The University of Southampton is phasing Turnitin out, which tells you the unease is not confined to student unions.
What Turnitin says
The company's position has two parts, and the distinction between them is where the whole argument sits. It says no update will take effect before 1 September 2027, and it says plainly that it does not use customer or student work to train the AI assistant inside its Clarity product.
But it also says it may use anonymised student submissions to improve its detection and assessment tools. That second sentence is the one to read twice. Anonymised means your name is removed; it does not mean your writing is. And "improve our detection tools" is, in practice, machine learning on a corpus of student essays — the distinction the company is drawing is between training a chatbot and training a classifier, which is a real technical difference that will not feel like one to the person who wrote the essay.
The wider problem it sits inside
Plagiarism detection was always a slightly awkward bargain: students supply the archive that makes the product work, and the product is sold back to their institution. That was tolerable while the archive only did one job. Now that a corpus of millions of human-written, graded, dated essays is among the most commercially valuable datasets in existence — precisely because it predates and can be used to identify machine-written text — the bargain is worth revisiting.
What it means in Bangladesh
Bangladeshi universities use the same software. Turnitin reaches public universities through UGC arrangements and is a standard requirement for theses and dissertations at institutions like DU, BUET, NSU, BRACU and IUB, usually with a hard similarity cap before a thesis can be defended. Every one of those documents — original research by Bangladeshi students, in English and in Bangla — is in the same archive, under a licence the students never read and could not have refused.
Three practical steps are available now. Universities should ask their vendor, in writing and before the next renewal, what may be done with submitted work beyond similarity checking, and keep the answer on file. The UGC should decide whether a national licence permits secondary use at all, because negotiating that once at the national level is worth more than 40 institutions each negotiating it badly. And departments should tell students plainly, at submission time, where their work goes and how long it is kept. We have written before that institutions here are not ready for AI; this is what unreadiness looks like in a contract rather than a classroom.
