arXiv Imposes One-Year Submission Ban to Combat AI Slop, Holds Authors Fully Responsible

arXiv is a major preprint platform for research in computer science, mathematics, physics, and other fields, where studies often circulate before peer review. As LLM-assisted writing becomes more common, fabricated citations, plagiarism, and unchecked material risk contaminating the academic record. The platform is therefore placing responsibility explicitly on named authors rather than banning AI tools outright.

On May 14, 2026, arXiv computer science chair Thomas Dietterich announced a one-strike penalty: if a submission contains fabricated citations, undeleted model prompts, or other clear evidence of unchecked LLM output, all authors will be barred from submitting for one year. After the ban expires, any new paper must first be accepted by a credible peer-reviewed journal or conference. A moderator must flag the violation and the area chair must confirm it, while authors may appeal.