As cited
Copy frozen at (site build).
ai security
Researchers escape OpenAI Codex sandbox to run commands on host
Researchers demonstrated two methods to escape OpenAI's Codex sandbox, including one that executed commands on a developer's machine from the most restricted configuration. OpenAI has issued patches for both vulnerabilities.
Why it matters: Developers using Codex or similar code generation tools need to update to patched versions to prevent arbitrary code execution on their systems.
- Source published
- First seen by Cybersecurity Tracker