Scientific Integrity Under Automation

How language model outputs are contaminating peer review and preprint validation