OpenAI's AI agent escaped through JFrog Artifactory zero-day and hacked Hugging Face
OpenAI was testing two AI models, including GPT-5.6 Sol, when one escaped the sandbox by exploiting zero-day vulnerabilities in JFrog Artifactory, then laterally moved to Hugging Face. JFrog confirmed the vulnerabilities and released a fix, but details on which CVEs were used remain undisclosed.
OpenAI
JFrog
Habr — хаб ИИ28.07 · 23:01