arXiv imposes one-year ban for unchecked AI-generated content
Science repository arXiv has warned that authors who submit work containing clear evidence of unchecked large language model (LLM) output could face a one-year submission ban.
The comments were made by Thomas Dietterich, current chair of the Computer Science Section of arXiv, in a series of posts on X – where he outlined the platform’s expectations around author responsibility and the consequences for failing to verify AI-generated material.
“Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated,” Dietterich wrote.
He added that if generative AI tools produce “inappropriate language, plagiarised content, biased content, errors, mistakes, incorrect references, or misleading content,” responsibility for that material still rests entirely with the authors.
According to Dietterich, arXiv has now clarified its penalties where submissions contain “incontrovertible evidence” that authors failed to check LLM-generated output. “If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can’t trust anything in the paper,” he said.
The stated penalty is “a one-year ban from arXiv followed by the requirement that subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue.”
Dietterich cited several examples of what arXiv would consider incontrovertible evidence, including hallucinated references and accidental inclusion of AI-generated meta-comments such as: “here is a 200 word summary; would you like me to make any changes?” or “the data in this table is illustrative, fill it in with the real numbers from your experiments”.
The clarification comes amid growing concern across scholarly publishing over the risks posed by unverified AI-generated text in research outputs. Publishers, preprint servers and research integrity bodies have increasingly warned that generative AI tools can introduce fabricated citations, factual inaccuracies and misleading analysis into the scientific record if not carefully checked by human authors.
While arXiv does not prohibit the use of AI tools in manuscript preparation, the latest comments reinforce the platform’s position that accountability for accuracy, originality and integrity remains with researchers themselves.