All OpenReview Data Leaks (twitter.com)

Reports indicate a comprehensive leak of OpenReview data — the platform widely used for conference submissions and peer review — exposing submissions, reviews, metadata and potentially reviewer and author identities. If verified, this undermines double-blind review norms, risks exposure of unpublished work and sensitive critique, and could enable harassment, plagiarism, or legal/privacy claims. The incident matters to the AI/ML community because OpenReview houses a large corpus of cutting‑edge research and confidential peer-review conversations; leaked content can be copied into training sets, used to deanonymize participants, or weaponized to manipulate future reviews and reputations. Technically, leaked items reportedly include PDFs, review text, timestamps, emails and other metadata; common root causes for such breaches are misconfigured storage (e.g., public S3 buckets), exposed API keys, or broken access-control on databases/search indices. Immediate implications include the need to audit logs, rotate credentials and tokens, remove public copies, and notify affected individuals and conferences. Longer-term risks are contamination of public datasets used for model training, erosion of incentives for candid peer review, and legal/regulatory exposure. Communities should treat affected content as potentially public, push for stronger platform security and anonymization practices, and coordinate incident response across conferences and institutions.