Good question Salvatore. I think it's unfair to living artists that these wealthy companies are (in part) using their work to train models they will then charge the median artist for using if they want to stay relevant (among the high-quality ones Stable Diffusion is an exception because it is open-source).

One solution is to ask them first before training a new model if they want to participate. Or, at the very least, companies should retribute them after the fact or add an opt-out feature.