Security researchers found two ways out of the OpenAI Codex sandbox, one of them capable of running commands on a developer's machine from Codex's most locked-down mode, with no approval prompt and ...

Codex’s network proxy was resolving a hostname before it checked whether that host was allowed, so the name went out to ...

Two OpenAI Codex sandbox flaws, Overpatch and Heapjack, could let malicious repositories execute commands on developer systems.

Security researchers broke out of the sandboxes in four widely used AI coding agents, including Cursor, OpenAI's Codex, Google's Gemini CLI and Antigravity, without attacking the sandbox head-on. The ...