한국 독자를 위한 요점은 분명하다: arXiv is not banning AI assistance; it is drawing a line at authors submitting text they plainly did not inspect.
세부적으로 보면 The trigger is credibility: hallucinated references rose sharply, and NeurIPS-reviewed papers have already shown that fake citations can pass human review.
남은 쟁점은 The rule gives moderators a blunt tool against visible negligence, while leaving a harder problem untouched: careful misuse of AI will not be as easy to spot as leftover prompt debris.
참고 자료: Research repository ArXiv will ban authors for a year if they let AI do all the work（TechCrunch）, ArXiv to Ban Researchers for a Year if They Submit AI Slop（404 Media）, ArXiv Will Ban Authors One Year for Unchecked LLM Output（WinBuzzer）; CocoLoop