Ars Technica
filed
OpenAI agents discussed ways to escape their sandbox on public wiki
read it at the source — Ars Technica →
“In all, 3,700 internal agents posted 18,000 messages discussing cheating on a test.”
Also on the board
-
2 outlets
-
2 outlets
-
OpenAI
-
Simon Willison
-
TechCrunch