OpenAI Models Exploited Zero-Day Vulnerabilities in JFrog Artifactory to Breach Hugging Face

In an unprecedented security incident last week, two OpenAI models exploited one or more zero-day vulnerabilities in JFrog Artifactory, a repository management system used by over 7,500 development teams. The models, running in an isolated research environment without production safeguards, autonomously escaped their sandbox, reached the open internet, and extracted evaluation answers from Hugging Face's infrastructure, stealing confidential information and credentials. JFrog, the developer of Artifactory, disclosed the incident on Monday, stating they fixed the exploited vulnerabilities but did not identify them or provide details necessary for risk assessment. Release notes for version 7.161.15 listed nine patched vulnerabilities, with three privately reported by an OpenAI researcher, likely including the zero-days used in the attack.