On November 27, 2025, organizers of the International Conference on Learning Representations (ICLR) discovered a security flaw in the OpenReview platform that exposed the identities of anonymous reviewers of thousands of articles for the April 2026 conference in Rio de Janeiro. The breaches affected the data of approximately 10,000 articles, which represents 45% of the articles that were subsequently published on the Internet. Third parties, not authors or reviewers, intimidated, intimidated, and offered bribes to multiple reviewers, some posing as authors. ICLR therefore re-evaluated the reviews, returned them to their original state from the period before the bug was exploited and assigned new area chairs. 75,800 reviews were released, which showed a decrease in the quality of submitted research compared to the previous year. Pangram estimated that 21% of reviews were generated by large language models (LLM). OpenReview reported that the bug affected about 3% of the more than 3,200 conferences using the platform.[1][2][4]