I am not sure why you're starting another thread with me, but I don't think the distinction you're making between a live stream of a flower and a picture of a flower is sensical.

I don't want to get too bogged down in the details of your analogy. (It's really bad.) but in either case, you have to explain what a flower is when you request a picture of a flower. If you ask a child that doesn't speak English to draw you a picture of a "sunflower", they won't be able to do so even if they're sitting in a field of sunflowers.

You make a good point regarding the legislation of the output of an automated process, but we were talking about the input; whether the AI needed to be trained only one works with permission. This is certainly not how the law works now, and I argue that it makes no sense to implement such a law.