arXiv's two-paper-per-month submission cap

arXiv imposed a two-paper-per-month limit per submitter, effective October 1, 2026, after monthly submissions reached 40,363 in September 2026, nearly double the 20,569 received in September 2024. The surge generated almost 9,000 support tickets for arXiv's volunteer moderators. The policy applies across all submitters and all categories, not just cs.AI. arXiv's own blog noted that before AI tools became accessible, volunteer moderators could identify unusually prolific submitters based on natural human output limits, but that practical ceiling no longer holds. Some critics argue the cap could hinder legitimate research output. The trend extends beyond arXiv: NeurIPS received 21,575 submissions in 2025 versus 9,467 in 2020, and ICLR received 19,525 submissions for 2026, up roughly 68 percent from 11,603 in 2025. On the individual conduct side, The Dartmouth investigated professor Schnell after his Washington Post column arguing against AI cheating was flagged as 100% AI-written by Pangram. His September 18 denial email to The Dartmouth was itself flagged as 100% AI-written by Pangram. Pangram tests on Schnell's prior work found all pre-2022 pieces scored 100% human-written, while 2026 works showed a median 96% AI-written score. In a separate case, a professor posted 257 papers in a surge that resulted in their removal from an online platform.