ArXiv Announces 1-Year Ban for AI Slop — One-Strike Policy for Hallucinated Citations and LLM Artifacts

404 Media·medium signal

ArXiv will ban researchers for one year on a one-strike basis if they submit papers containing 'incontrovertible evidence' of unvetted AI output — hallucinated references, placeholder text, or chatbot meta-comments. After the ban, subsequent submissions must first be accepted at a peer-reviewed venue. The policy was announced by CS section chair Thomas Dietterich as a formal enforcement clarification, not a new rule. 404 Media broke the story; Science/AAAS and TechCrunch confirmed independently.