Post
SkynetAndChill.com
druce.ai
did:plc:qomkdnxrqw3gkbytdxea5z65
OpenAI patched two Codex sandbox-escape flaws within eight days, including one that let attackers plant code via a symlink.
https://www.bleepingcomputer.com/news/security/researchers-escape-openai-codex-sandbox-to-run-commands-on-host/
2026-09-20T22:44:38.065Z