arXiv, one of the world’s most widely used repositories for scientific preprint research, has announced stricter measures aimed at curbing the careless use of large language models in academic papers as concerns grow over the spread of low quality AI generated research.
The platform, which plays a major role in circulating research findings in fields such as computer science and mathematics before peer review, said authors who submit papers containing clear evidence of unchecked AI generated content could face sanctions. According to Thomas Dietterich, chair of arXiv’s computer science section, evidence such as hallucinated references or visible comments exchanged with AI tools would indicate that authors failed to properly verify their work.
Under the new policy, affected authors risk a one year ban from the platform and may also be required to publish future submissions in reputable peer reviewed journals before they can return to arXiv. Dietterich explained that the rule is not a ban on the use of artificial intelligence tools, but rather an insistence that researchers take full responsibility for all content included in their papers, regardless of how it was produced.
The development comes amid growing concerns within the academic community over fabricated citations and misleading AI generated content appearing in scientific literature. arXiv, which is transitioning from Cornell University into an independent nonprofit organisation, said the move forms part of broader efforts to strengthen research integrity and tackle the increasing volume of AI generated academic material flooding online repositories.