According to subject matter experts and insiders at the National Testing Agency (NTA), the core of the dispute surrounding UGC-NET lies in the 'reckless use of AI' during the preparation and translation of examination materials, as well as an unexplained six-week delay in canceling three flawed tests. Consequently, NTA was forced to schedule re-examinations for English, Commerce, and Sociology.
However, the testing agency itself categorically rejected the claim that generative AI was used to create the controversial materials. Experts told TOI on condition of anonymity that AI was widely employed in preparing these tests. Former experts previously associated with NTA asserted that AI was used not only for drafting UGC-NET questions but also for translating NEET-UG materials. They attributed the scale and nature of the errors to insufficient human oversight.
This incident has drawn attention to the depth to which NTA has integrated artificial intelligence into the process of preparing examination materials, particularly in translation. A senior source within the Ministry of Education revealed that they held several meetings with two high-ranking NTA officials who admitted they could not release answer keys because the tests were developed 'ad hoc,' to which, when asked what that meant, they replied 'nothing.' When asked why experts did not check the test items, the source was unavailable.
The agency's reliance on technology itself is not proof that the errors in the canceled exams were caused by AI. Another source noted that a panel of four to five experts usually develops a test paper. This source questioned how numerous questions from previous years' tests were selected, and how examiners could be unaware of the answers to questions they themselves designed.
Nevertheless, NTA Director General Abhishek Singh refuted the use of generative AI in creating the disputed materials. He told TOI that 'human review' was conducted after the tests were compiled. When asked how errors could have gone unnoticed, he could add nothing new.
NTA's reaction upon discovering discrepancies has also come under scrutiny. Officials from the Ministry of Education and sources within NTA reported that the ministry advocated for the cancellation of the tests and the commencement of re-testing by the end of July or the first week of August. However, NTA did not follow this advice, waiting nearly six weeks and announcing the cancellation only after the preliminary answer keys for 84 out of 87 subjects were published.
The Sociology exam drew severe criticism. Candidates complained about incorrectly written names of world-renowned scientists, distorted terminology, clumsy translations, and the fact that the questions bore little resemblance to the prescribed curriculum. It was reported that George Ritzer was listed as 'Putzer,' Talcott Parsons as 'Parsow,' G. S. Ghurie as 'Ghunye,' and Martha Nussbaum's name was also distorted.
The English exam raised a separate issue related to the mass repetition of material. Candidates claimed that about 67 out of 150 questions appeared in the previous UGC-NET exam, with some stating that even the sequence of answer choices was preserved. Commerce candidates also reported the repetition of several questions from old exams, including those concerning taxation, break-even analysis, capital structure, and service marketing. How these repetitions occurred remains undetermined.
A lecturer from JNU stated that 'the bigger issue is the extent to which automated systems are used for translating, organizing, and assembling national high-level examinations and whether adequate human academic review was conducted after this process.' Since UGC-NET determines eligibility for academic and research careers in Indian universities, the reliability of the examination process is particularly crucial.
New statements deepen the debate beyond poor test construction. They call into question who authorized the use of AI, what tools were involved, whether qualified experts reviewed every question, and why NTA delayed the cancellation despite the ministry's opinion. The central question is no longer just how many errors slipped into the three national exams. Experts believe the issue is what part of the examination process was delegated to machines and whether enough qualified people reviewed what the machines produced. For an exam that determines eligibility for teaching and research careers, the question is whether automation is permitted to replace academic judgment and responsibility.
