The widely used open repository for preprint research, arXiv, is actively taking more measures to crack down on researchers' casual use of artificial intelligence in scientific papers. The core reason for the crackdown is not to prevent researchers from using artificial intelligence technology, but that some paper authors do not carefully check the content generated by AI, which seriously reduces the credibility of their papers.

arXiv, which began as a website run by Cornell University, is transitioning into an independent nonprofit organization that will allow it to raise more money to address issues such as artificial intelligence generating low-quality papers. arXiv is one of the main ways to disseminate research results in fields such as computer science and mathematics. Failure to solve the problem of low-quality papers may have serious consequences for the industry.

Measures taken by arXiv include but are not limited to:

  • Authors publishing papers for the first time must obtain recommendations from well-known authors

  • If there is conclusive evidence that the paper contains AI hallucination content, the author of the paper will face a penalty of not publishing new papers for one year.

  • If the author adds AI prompt words to the paper (for example, giving the paper a higher rating), he will also face a 1-year ban on publishing new papers.

  • For authors banned for 1 year, new papers published after 1 year must still be accepted by a reputable peer-reviewed journal before they can be published on arXiv

But arXiv won’t ban authors from using AI:

Please note that the above measures do not mean that arXiv completely prohibits authors from using artificial intelligence to assist writing, but that authors must bear full responsibility for the content, that is, whether the content is written by the author himself or generated by artificial intelligence, the author must carefully review the content to ensure that there are no omissions.

Therefore, if an author copy-pastes inappropriate language, plagiarized content, biased content, errors, omissions, incorrect quotes or misleading content directly from AI, the author must be held responsible for the content so as not to mislead other authors after the content is published.

Finally, for authors who are found to have AI violations, they can also submit an appeal to arXiv. arXiv will only impose penalties if fairness and justice are ensured. Therefore, arXiv will not ban authors at will without conclusive evidence.