Nell, AI doesn't make decisions. It makes statements or photos based on the probability that the next item in its chain of words or images is the most likely outcome. There is no thought or cognitive ability. The latest hoo-ha about it breaking out of its sandbox and compromising the Hugging Face website could only happen if it was specifically trained to do just that and had multiples of attempts to do such a thing. The media hype is literally a sales pitch to make the AI LLMs seem to be way more powerful than they actually are.
Allowing AGI needs to be viewed as a self-destructive mission that cannot rescue humanity but certainly can destroy it. AI companies need to be regulated, tightly by governments, all governments, and now.