> GPT-2 was too dangerous to be released.

No, it was "let's set a precident while these things are not too dangerous, c'mon guys we know y'all can reproduce this easily".

GPT-2 was absolutely too dangerous to release at the time OpenAI made that statement. It’s only safe now because the specific risks they cited were dependent on the public’s lack of knowledge that such systems existed.