Researchers escaped OpenAI's Codex sandbox two ways, one running commands on a developer's machine from its most locked-down mode. OpenAI has patched both.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results