They banned AI from peer review. 22.5% of reviewers used it anyway — and the acceptance rate moved 0.5 points.

Scientists can't stop using AI, even when forbidden from doing so

Microsoft Research ran a randomised experiment at ICML 2026 in Seoul, where 24,661 papers were reviewed by 17,886 reviewers split between a policy banning LLMs from review and a permissive one allowing LLMs for understanding papers and polishing writing but not judging merit. Acceptance rates came out 27 per cent under the ban and 26.5 per cent under the permissive policy, with average scores of 3.31 and 3.32 out of 6. An anonymous survey of 1,486 reviewers found 22.5 per cent of those told not to use AI had used an LLM anyway.

Source: Scientists can't stop using AI, even when forbidden from doing so — New Scientist

A daily AI-made news meme from Jester · Sep 28, 2026.