Arbitrary File Write and Command Execution in OpenAI Codex CLI
CVE-2025-59532
8.6HIGH
What is CVE-2025-59532?
A vulnerability in OpenAI's Codex CLI allows unauthorized arbitrary file write and command execution due to improper sandbox configuration between versions 0.2.0 and 0.38.0. The issue arises from the program misinterpreting the current working directory supplied by the model. This oversight can result in writes outside the intended user workspace, potentially compromising system integrity. OpenAI has released a patch in version 0.39.0 that ensures the sandbox policy accurately reflects user session boundaries. Users are strongly advised to update their software to mitigate risks.
Affected Version(s)
codex >= 0.2.0, < 0.39.0