Major conference catches illicit AI use — and rejects hundreds of papers

2026-03-25 · Source: Machine learning : nature.com subject feeds · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, quick

Summary

The International Conference on Machine Learning (ICML), scheduled for July in Seoul, rejected 497 papers, approximately 2% of submissions, due to authors violating AI-use policies during peer review. ICML organizers implemented a watermarking system in research papers distributed for review. This system embedded hidden instructions that prompted large language models (LLMs) to include specific phrases in generated reviews, thereby revealing AI assistance. The conference has a reciprocal review policy, requiring authors to review other papers. Authors whose reviews were found to be AI-generated had their own submissions rejected. This action, announced in a March 18 blog post, aims to protect trust within the rapidly evolving AI research community. Despite some researchers applauding the move, others, like Zhengzhong Tu, a computer scientist at Texas A&M University, expressed concerns that such policies might demotivate reviewers and lead to the generation of "meaningless reviews" by LLMs.

Key takeaway

For AI Scientists involved in peer review, this ICML action highlights the increasing scrutiny on AI-assisted review processes. You should be aware that conferences are deploying sophisticated methods, like watermarking, to detect policy violations. Ensure your review practices strictly adhere to stated LLM-use policies, even if they seem restrictive, to avoid potential rejection of your own research submissions and maintain professional integrity within the community.

Key insights

Watermarking can detect AI-generated content in peer review, leading to policy enforcement and paper rejections.

Principles

• Protect community trust in rapidly changing fields.
• Clear guidance is needed for responsible AI use.

Method

ICML embedded watermarks in review papers that prompted LLMs to include specific phrases if used, enabling detection of AI-generated peer reviews.

In practice

• Implement watermarking for content integrity checks.
• Establish clear LLM-use policies for submissions.
• Consider dual review streams for policy flexibility.

Topics

• ICML
• Large Language Models
• Peer Review
• AI Ethics
• Digital Watermarking

Best for: AI Scientist, AI Researcher, Research Scientist, AI Ethicist

Related on AIssential

• [D] ICML paper to review is fully AI generated — AI-generated research papers challenge conference policies and reviewer assessment of technical merit and authorship.
• [D] ICML 2026 Review Discussion — ML peer review suffers from high variance in reviewer quality and increasing submission volumes.
• A key science publishing platform is cracking down on AI slop — arXiv's ban on AI-error papers highlights the challenge of maintaining research quality amidst rising AI-generated content.
• PeerPrism: Peer Evaluation Expertise vs Review-writing AI — Current LLM detectors conflate text generation with intellectual contribution in hybrid human-AI peer review.
• Can AI Review Improve Paper Drafting? An Empirical Study on 20 Computer Architecture Submissions — AI review can significantly align with human feedback and identify novel issues in paper drafting.
• The AI Scientist: Towards Fully Automated AI Research, Now Published in Nature — AI agents can now autonomously conduct machine learning research, generate papers, and pass human peer review.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.