My understanding is this was still a bit of a grey area: particularly with non-text media?

I thought that the training would not be covered but that there is the possibility of LLMs regurgitating the training materials under certain circumstances which would be covered as a potential breach?

Even without consideration of AI though I still think its an important question. Do users retain the copyright of thier work? I don't want to see another repeat of other platforms where users contribute and build communities with a collective mindset giving the platform it's value only to be enshittified.