Additional clarification because of course my brain won't shut off that easily:

The main problem with what's being called AI today is that it's based upon all the input those running the machines could lay their hands upon. There's some factual stuff in there, but it's not only extremely minimal when compared to the whole, but what's factual is effectively impossible for machines to distinguish from the rest. Humans are supposed to help them with that, but — as strongly exemplified by Elon Musk effectively destroying the once very promising Grok AI with his manipulations — the quality of that "help" can be extremely suspect.

Genuine AI will need to be fed strictly facts as best the collective human race has been able to determine. It will also need a very strong foundation of human psychology (and likely politics & religion) in order to have a chance of catching any bad inputs that still slip through.

Without a solid foundation, genuine AI will never happen. Whether humans will be able to put aside their bullshit in time to achieve this honestly doesn't look hopeful.