[0xpgm]

If I remember correctly, the original GPT was considered too dangerous to release to the wild.

With hindsight, does that hold? If not, then how would we know a model is truly dangerous to release?