Re:OpenAI Model Misalignment Report
by RandomLensman on Thursday September 17, 2026 @07:49AM
(#49739420)
Could sit in the box and interact if a model of certain capabilities is needed/tested. We do physical security for other things, too. Not saying everything needs that type of isolation.
[ Parent
| on HN ]
There are no comments on this comment yet.