arXiv caps submissions at two per person per month as AI paper flood hits moderators

Starting October 1, 2026, the preprint platform arXiv will allow only two submissions per person per calendar month. The existing cap of three active submissions at any given time, in place since 2024, stays. According to arXiv, the limit applies across all categories and only to the person who submits the paper, so co-authors are not affected. Rejected papers still count toward the limit because they use up moderation time. Authors who delete a paper before publication get that slot back.

arXiv points to a sharp rise in submissions, partly driven by AI tools, as the reason. The platform received 9,869 submissions in September 2016, 20,569 in September 2024 and 40,363 in September 2026. That surge generated nearly 9,000 support tickets for staff and moderators, and submissions in the cs.AI category alone grew more than sixfold in two years. Volunteer moderators are seeing more thin, narrowly focused papers, "salami" papers sliced into tiny pieces, and dense texts written by AI.

Thomas G. Dietterich, chair of the arXiv Editorial Advisory Council, calls the volunteer moderators the backbone of quality control. He writes: "However, a relatively small proportion of authors are submitting a large number of low-quality papers and consuming a disproportionate fraction of the moderators’ time." He adds that this is unfair to authors who keep submitting quality papers, whose work can be delayed for days or weeks. On Bluesky, he says some authors submit dozens of papers.

Dietterich also hopes the limit will curb "micro-result" papers, for example one that describes a single extra ablation study or a minor tweak to an architecture. arXiv calls the rule a temporary fix until its own moderation tools can scale, and Dietterich says he hopes the cap can be raised or dropped later.

This is the latest of several tightenings. In November 2025 arXiv tightened rules for computer science papers, and since then review articles and position papers must go through peer review. In May 2026 Dietterich announced a one-year ban for authors whose papers clearly contain unverified LLM outputs such as hallucinated sources. AI conferences face the same pressure: ICLR 2027 had received about 50,000 abstracts before its deadline, while ICLR 2026 got around 19,500 valid submissions. Among roughly 5,000 accepted papers at NeurIPS 2025, GPTZero found at least 100 fabricated citations, even though each paper had at least three reviewers.

The change lands as arXiv goes independent. After 25 years, it is splitting from Cornell University and has been transitioning to an independent nonprofit since July 1, 2026. In late September, the Simons Foundation International, XTX Markets and the Siegel Family Endowment pledged a total of $17.2 million, spread over three to five years. arXiv plans to put part of the money toward technical upgrades, with a focus on handling AI-generated content.

Meanwhile the major AI labs are pushing to position their models as research tools. In August, Google Deepmind expanded its Co-Scientist into a system designed to plan experiments, control lab equipment and generate manuscripts. Anthropic is building its own biology lab where Claude is meant to guide robots through experiments. In September, OpenAI announced an "automated research intern" that handles defined research tasks under human oversight. A study involving Princeton and the UK AI Safety Institute took a different angle: it had Claude Opus 4.8 work for six days on research questions from two unpublished NeurIPS papers, and the original authors, acting as reviewers, rejected both sets of results.

Key facts

From October 1, 2026, arXiv allows only two submissions per person per calendar month, across all categories; the existing cap of three active submissions at a time stays.
Monthly submissions rose from 9,869 in September 2016 to 20,569 in September 2024 and 40,363 in September 2026, generating nearly 9,000 support tickets for staff and moderators.
Rejected papers still count toward the limit, but deleting a paper before publication returns the slot; co-authors are not affected.
arXiv calls the rule a temporary fix until its moderation tools can scale, and Thomas G. Dietterich hopes the cap can be raised or dropped later.
arXiv is becoming an independent nonprofit and received $17.2 million in pledges over three to five years, part of it earmarked for handling AI-generated content.

Why it matters

arXiv is a central venue for sharing AI and computer science research, and it is now rationing access by volume. The numbers show why: submissions in September 2026 (40,363) were nearly double those of September 2024 (20,569), and cs.AI alone grew more than sixfold in two years. The cap is also one more step in a series of tightenings, after the November 2025 peer-review requirement for review articles and position papers and the May 2026 one-year ban for unverified LLM outputs. AI conferences report similar strain, with ICLR 2027 receiving about 50,000 abstracts before its deadline.

Who it affects

Anyone who submits papers to arXiv. The limit applies to the person who submits the paper, not to co-authors, so a group can spread submissions across its members. Volunteer moderators, who the article says are seeing thin, "salami" and AI-written papers, are the people the rule is meant to relieve. Authors who regularly submit quality papers are described by Dietterich as the ones currently harmed by delays of days or weeks. Dietterich says some authors submit dozens of papers, and those are the ones the cap bites hardest.

How to use it

Plan around two submissions per calendar month from October 1, 2026, across all categories. The existing cap of three active submissions at any given time still applies. A rejected paper still uses one of the two slots, so submit work that is ready. If you delete a paper before publication, you get the slot back. Co-authors are not affected, because the limit counts only the person who submits.

How solid is it

The rule is an arXiv policy announced by Thomas G. Dietterich, chair of the arXiv Editorial Advisory Council, and the submission counts and ticket figures come from arXiv as reported by The Decoder. arXiv attributes the growth partly to AI tools; the share of the 40,363 September 2026 submissions that is AI-generated is not stated. The conference figures (ICLR, NeurIPS) and the GPTZero finding of at least 100 fabricated citations among roughly 5,000 accepted NeurIPS 2025 papers are context reported alongside, not part of arXiv's announcement.

Risks and caveats

arXiv itself calls the cap a temporary fix until its own moderation tools can scale, and Dietterich hopes it can be raised or dropped later; no date for any review is given. How the cap will be enforced technically, and how a "person" is identified, is not stated. Because rejected papers count, authors lose a slot even when moderation turns a paper down. Dietterich says only a relatively small proportion of authors cause the problem, so the cap applies to everyone to address the behaviour of a few. The hope that it will curb "micro-result" papers is his expectation, not a measured outcome.

“However, a relatively small proportion of authors are submitting a large number of low-quality papers and consuming a disproportionate fraction of the moderators’ time”

— Thomas G. Dietterich, chair of the arXiv Editorial Advisory Council