True...but so is Google, right? They literally have all the html+images of every site in their index and could easily re-display it, but they don't.
Ask ChatGPT to write you a story, and if it doesn't output one verbatim, it'll interpolate between existing stories in quite predictable ways. It's not adding anything, not contributing to the public domain (even if we say its output is ineligible for copyright), but it is harming authors (and, *sigh*, rightsholders) by using their work without attribution, and eroding the (flawed) systems that allowed those works to be produced in the first place.
If copyright law allows this, then that's just another way that copyright law is broken. I say this as a nearly-lifelong proponent of the free culture movement.