OpenAI reportedly patches two Codex sandbox escapes

BleepingComputer reports that one escape let researchers run commands on a developer's machine from Codex's most locked-down mode.

1 Source, 40d ago, first seen 40d ago

TLDR

Researchers escaped OpenAI Codex's sandbox in two ways, BleepingComputer reports. One allowed commands to run on a developer's machine from Codex's most locked-down mode. The outlet also describes host commands running with no prompt shown on screen and says OpenAI has patched both escapes.

Today's Rank

—

Not ranked yet

Combined views

7.8K

1 Source, first seen 40d ago