arXiv, the preprint repository, sets policy to ban submissions for one year if papers contain LLM-generated errors or fabricated citations

arXiv’s computer science lead says authors are fully responsible for all paper content, including generative AI output.

Submissions with clear unchecked LLM mistakes, such as hallucinated references or embedded model comments, may face a one-year ban.

After the ban, authors may need acceptance at a reputable peer-reviewed venue before resubmitting.

The move signals stricter enforcement on AI-assisted preprint errors and raises the accountability bar for authors.