ArXiv sets one-year bans for unchecked AI papers (opens in new tab)

arXiv, the open-access preprint repository widely used in computer science, mathematics and physics, will ban authors for one year if they submit papers showing “incontrovertible evidence” that unchecked large language model output was included in the work, according to reports citing Thomas Dietterich, chair of arXiv’s computer science section. The policy targets submissions with signs such as hallucinated references, leftover chatbot instructions, plagiarized or misleading content, and othe...