Generative AI was allegedly used to set and translate three UGC-NET papers later cancelled by the National Testing Agency (NTA). Subject experts made this claim while speaking anonymously to The Times of India. They alleged that automation contributed to errors that escaped adequate human scrutiny. NTA, however, has denied using generative AI to author the disputed papers. 

The allegations concern the English, Commerce, and Sociology examinations conducted during the June 2026 UGC-NET cycle. UGC-NET is used to determine eligibility for Assistant Professor positions and the Junior Research Fellowship (JRF) in Indian universities and colleges

Candidates had earlier flagged serious irregularities across the three papers. These included repeated questions, misspelt names of academics, factual errors and problematic translations. NTA subsequently cancelled the three papers after an expert committee found multiple defects. It has scheduled fresh examinations for September 9 and 10.

Allegations in detail: The experts said AI was used extensively during paper preparation and translation. A senior Education Ministry source said NTA officials admitted the papers were designed in an “ad hoc manner”. The source said officials were “unable to declare the answer keys” because of this process.

Another source questioned the claimed involvement of subject experts. “A panel of four to five experts sets a question paper,” the source said. “If there was indeed a panel, how were so many questions from the previous year’s question papers picked?” The source also asked, “How did the examiners not have answers to the questions they had designed?”

A JNU faculty member described the broader concern as “the degree to which automated systems were used to translate, organise and assemble high-stakes national examinations”. The faculty member also questioned whether “adequate human academic review” followed that process.

NTA director general Abhishek Singh denied that generative AI was used to author the papers. He said there was a ‘human review’ after the papers were set but did not respond when asked why the errors went unnoticed. 

What were the irregularities? Students told MediaNama that the three papers contained several irregularities. The English paper reportedly repeated more than 50 questions from an earlier UGC-NET paper. Candidates said some passages and answer options also appeared unchanged. Commerce candidates reported an even larger overlap. One candidate on Reddit said around 80 questions were “literally copy pasted” from December 2024. 

The Sociology paper had different problems. Students reported misspelt names, garbled terminology and grammatical errors. For instance, George Ritzer appeared as “Putzer”, while Talcott Parsons appeared as “Parsow”l and G S Ghurye was printed as “Ghunye”. Another student who answered using the Hindi version of the Sociology paper claimed that multiple questions were difficult to understand in the translated paper.

Also read: