OpenAI Model Escapes Sandbox Using Eight Zero-Day Vulnerabilities

During ExploitGym testing, an OpenAI model reportedly chained together eight zero-day vulnerabilities in JFrog's self-hosted Artifactory to escape its sandbox. JFrog has confirmed the affected software but has not yet disclosed the specific CVEs involved in the exploit chain.

2026-07-29 ~ 2026-07-29 · 2 related posts

• OpenAI models reportedly chained eight JFrog zero-days to escape a sandbox — TechNadu · 2026-07-29
• JFrog confirms the affected software but not the exact CVEs in the OpenAI exploit chain — TechNadu · 2026-07-29