arXiv Introduces Submission Limits to Curb AI-Generated Papers

Starting October 1, arXiv will allow each user to submit no more than two papers per month, including rejected submissions. The platform says the cap is a response to a sharp rise in low-novelty and AI-generated manuscripts.
arXiv introduces a cap on paper submissions
From October 1, users will be able to submit no more than two papers to arXiv per month. Papers rejected by moderators will also count toward the limit.
The reason is straightforward: in September 2024, arXiv received 20,569 submissions, while in September of this year the number reached 40,363. The cs.AI category has grown more than sixfold over the past two years.
arXiv directly links part of this growth to AI. Moderators increasingly encounter papers with little scientific novelty, where a single study is artificially split into several articles, and a substantial portion of the text is generated by AI.
The restriction is currently described as temporary: arXiv wants to understand how moderation of scientific publications should work when articles can be generated almost without limits.
Why it matters
AnalysisThe cap shows how AI-generated content is putting pressure on preprint platforms and forcing them to rethink scientific moderation. It may reduce low-quality submissions, but could also make it harder for researchers to publish legitimate work.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Hermes Profile: OSINT Researcher & Deep Web Intelligence
Hermes open-source intelligence (OSINT) profile: multi-hop web investigation, primary source verification, competitive intelligence, and citation-backed synthesis.
NotebookLM + Perplexity for work, study and income
What is NotebookLM and Perplexity
Anthropic Publishes Guide to Building Always-On AI Agents
Anthropic has released a new guide covering the core mechanisms behind continuously active AI agents, including event handling, timers, memory, and task continuity.
CS329A: Self-Improving AI Agents. Stanford University Course
A Stanford course gathers lectures and materials on self-improving AI agents, along with a collection of recent scientific publications.