Three problems from a legal perspective with what you're recommending:

1) the new creation from a prompt is based on work product (which may have been copyrighted!) from other humans

2) originality - the first element for copyright protection is that the work must be original. If someone is simply typing a generic prompt like "picture of a cat" - how is that original? Not all prompts should be treated the same.

3) "modicum" of creativity - in my previous example - "picture of a cat" - how does that meet the minimum threshold for the second element of a copyrightable work, that it display at least some small amount of creativity? In that case, the AI is basically doing 99% of the work.