For example, I often use AI quotes in Articles here for specific Questions:

 However, I would have concerns about publishing ANYTHING that was totally composed by AI.

At least when I create a question for AI it is something I can look at and read to see if it sounds right and is likely accurate in what it says. However, then if you published an AI book you might never know whether some way AI conveys information might be fatal to some or one reader along the way.

This is the main problem  of AI written books. Liability leading to possibly fatalities of readers of those books.

Also, an AI book might send a younger person off on a totally bad idea which has no basis in physical reality at all.

So potentially sending a young person with no real knowledge of the physical world and how it actually operates in real time could send this young person in literally any direction or to their deaths.

So, this is why reading an AI book is so much "Caveat Emptor" or "Let the Buyer Beware".

I think what I worry about is when books might be published (or already are) that are written by AI and people are not told they are written by AI. This could really be a problem regarding the physical or mental survival of some human beings young or old or whatever age they might be. 

It once again brings to mind the old Computer adage for computer programmers in the 1960s:

"Garbage in Garbage out" which basically means you cannot get a good result unless you have good information programmed correctly in the first place. 

So, I worry about Garbage in Garbage out in relation to AI now too because the same basic rule applies to AI as well.