arXiv limits monthly submissions as AI changes research publishing
A new cap changes how researchers schedule papers and share responsibility for AI assisted work.
arXiv introduced a limit of two submissions per calendar month on October 1, applying across subjects and submitters. Its announcement describes the measure as temporary while the repository adapts moderation to growing use of AI.
September brought 40,363 submissions, compared with 20,569 two years earlier. arXiv says AI tools are contributing to the increase and placing pressure on volunteer moderation. The existing limit of three active submissions remains.
The limit follows the submitter
Rejected submissions count toward the monthly allowance. Coauthors do not spend their allowance merely by appearing on a paper someone else submits. Work deleted before public announcement does not count. A submission still on hold can occupy an active slot without consuming the next month’s monthly allowance.
That makes submission planning a practical part of a research group’s workflow. Before a conference deadline, collaborators can check who is submitting which paper, whether earlier work remains active and whether the manuscript is ready. Waiting until upload day could leave little room to resolve a scheduling problem.
Authors retain responsibility for AI assisted work
The repository’s moderation policy requires disclosure of significant use of generative AI tools according to the field’s methodological standards. Human authors remain responsible for the entire paper, including incorrect references, misleading statements and other errors introduced by a tool. AI language tools should not be listed as authors.
arXiv also distinguishes moderation from peer review. Its checks concern whether a submission fits the repository’s standards and scholarly purpose. A paper appearing there has not, by that fact alone, passed the review process of a scientific journal.
More papers still need more judgment
The change puts a familiar tension into a concrete rule. Tools can accelerate drafting, while deciding whether a result is original and worth circulating still demands attention. The useful response for a research team is to preserve that review step before submission, with someone explicitly accountable for checking the evidence and citations.
Featured image is an original AI generated conceptual editorial illustration.
