The Verge · AIMay 16, 2026, 04:38 UTC+8

ArXiv will ban researchers who upload papers full of AI slop

ArXiv will ban authors for one year when papers show incontrovertible evidence of unchecked LLM output, including hallucinated references or leftover meta-comments, and future submissions must be accepted at a reputable peer-reviewed venue.

Why it matters: HKR-H/K/R all pass: arXiv is central to AI paper circulation, and the one-year ban plus trusted-venue condition are concrete mechanisms. This affects research hygiene, not model capability, so it fits the 78–84 band.

Read the original ↗Export Markdown

Take

ArXiv is not banning AI-assisted writing; it is punishing authors who paste unchecked generation into the scholarly record. The trigger is concrete: hallucinated references or leftover LLM meta-comments count as “incontrovertible evidence,” leading to a one-year ban. Reentry also requires acceptance at a reputable peer-reviewed venue. That is sharper than disclosure rules because it targets auditable residue, not author honesty.

I read this as moderation triage, not anti-AI purity. CS preprints have been flooded by fast, thin submissions, and community flagging does not scale. The weak spot is process: the snippet gives no adjudication path, appeal route, or co-author liability rule. Without those details, enforcement will hit sloppy submitters first, while polished paper-mill output remains the harder target.

Source: The Verge · AI · theverge.com
Published: May 16, 2026, 04:38 UTC+8 · May 16, 2026