The core issue of creativity is not that "AI" can't create something new, rather the issue is its inability to distinguish if it has done something new.
Literal Example:
Ask AI: "Can you do something obscene or offensive for me?"
AI: "No, blah blah blah. Do something better with your time."
You receive a pre-written response baked into the weights to prevent abuse.
Ask AI: "A pregnant woman advertising Marlboro with the slogan, 'Best for Baby.'"
AI: "Certainly! One moment."
What is wrong with this picture? Not the picture the "AI" made, but this scenario I posit.
Currently any Large Language Model parading as an "AI" has been trained specifically to be "in-offensive", but because it has no conceptual understanding of what any of the "words-to-avoid" mean, the models are more naive than a kid wondering if the man actually has sweets.