Story perspectives

ICML 2026 Rejects 497 Papers Over AI Review Violations

3/26/2026

Summary

Story summary

In 2026, the International Conference on Machine Learning (ICML) in Seoul rejected 497 papers, 2% of submissions, for violations of AI-use policies in peer reviews.
Organizers implemented a watermarking system to detect AI-generated reviews and bolster trust in the research community.
Some researchers supported ICML's actions, while others criticized the policy's effectiveness.
ICML introduced two peer-review streams—one permitting limited AI use and one prohibiting it—allowing authors and reviewers to choose.

1 of 1

Story summary

In 2026, the International Conference on Machine Learning (ICML) in Seoul rejected 497 papers, 2% of submissions, for violations of AI-use policies in peer reviews.
Organizers implemented a watermarking system to detect AI-generated reviews and bolster trust in the research community.
Some researchers supported ICML's actions, while others criticized the policy's effectiveness.
ICML introduced two peer-review streams—one permitting limited AI use and one prohibiting it—allowing authors and reviewers to choose.