I think that could be the case if a derivative work closely resembles existing work, but I’m not sure that’s the case. I would imagine that the vast majority of the art it’s trained on can be freely viewed by anyone on the internet, so it couldn’t really have much of an effect on the market.

Sufficiently transformative works are covered under fair use. See Cariou v. Prince.

Was the art illegally obtained? Obtaining the art is very different from using it.

For-profit entities can still be good for society. I don’t think the for/non-profit status of the entities training the model would really affect whether the creators of data in its training set are being injured.

Fair use protections still apply to for-profit activities.

Then you may need to demonstrate how your posts being in a training set causes a detriment to you. I think this may be a hard bar to clear when there’s an argument that it’s fair use and highly transformative.

I don’t think you are entitled to compensation for your posts.

What detriment am I incurring by having my posts in a training set?

Why would it be beneficial for someone for there to exist a precedent that their public posts can be excluded from being used to train LLMs?

What’s the point? Opting out will make very little impact on the model.