ArXiv Penalty for AI Slop Draws Researcher Criticism
The preprint server arXiv announced a new policy that bans authors for one year from submitting papers if their work contains clear errors generated by AI tools, such as fabricated citations. The policy, stated by arXiv computer science section head Thomas G. Dietterich on X, holds all co-authors jointly responsible for AI-generated mistakes. University of Adelaide education researcher Vitomir Kovanović offered a critical perspective on the move. Kovanović acknowledged that most researchers agree AI-generated text itself is not the problem, but rather the ease with which it can produce low-quality work. However, he argued the blanket one-year ban on all listed authors is out of step with modern research practices, where papers often have four to five authors and sometimes hundreds. He noted that no equivalent sanctions exist for other problematic content, such as spreading fringe theories or using poor evidence. Kovanović suggested that AI systems could instead be used to verify references and check statistical analyses, proposing that automated checks followed by human review might be a more effective path forward than harsh penalties for minor AI-related violations.