TechCrunch reports arXiv will impose one-year bans when submissions show clear evidence of AI-generated paper slop.