Comment by chrisjj

> Why should OpenAI (or any frontier lab) be building these systems if they can't get a secure environment / containment right?

The can, because they've lowered expectations to a level even they can meet.