arXiv Tightens Rules for LLM Authors: 1-Year Ban for Unverified Content

Update: arXiv is tightening controls on LLM submissions.

Key points:
- T. Ditterich: if "conclusive" evidence shows authors failed to verify LLM outputs (fabricated citations, LLM annotations), a submission is untrustworthy.
- Sanctions: 1-year ban; subsequent submissions only allowed after acceptance in a peer-reviewed journal. Violations are confirmed by moderators and section chairs; appeals are possible.
- LLMs are not banned — the central requirement is full author responsibility. arXiv will add verifications for newcomers and is becoming independent to combat low-quality material.

Why it matters: a signal that standards for scientific reliability in AI use are rising.

How do you assess this policy?

#AI #science #researchethics #arXiv