Preprint repository arXiv has instituted strict rate limits on paper submissions in response to an overwhelming volume of AI-generated content. Under the new policy, researchers are restricted to two submissions per calendar month and a maximum of three active submissions at any time, aiming to relieve strain on its volunteer moderation team.
The volume of monthly submissions to arXiv has exploded, rising from 9,869 in September 2016 to 40,363 in September 2026. Computer science submissions have sextupled over recent years, driven largely by authors using large language models to draft papers, write code, and run automated analysis, often yielding repetitive or poorly executed “salami” papers.
While arXiv policies permit AI assistance if disclosed and substantively valuable, moderators report an unsustainable rise in dense, low-quality submissions. The caps follow previous counter-measures, including bans on review articles, stricter author endorsement rules, and temporary bans for users submitting AI slop.
Why it matters
AI-generated academic spam threatens open-science infrastructure, driving platforms toward hard submission quotas.
Researchers face increasing friction and delays publishing legitimate computer science work due to automated peer reviews.
Verifying authenticity in synthetic media and academic output is becoming a critical challenge across open platforms.
Source: 404media.co



