Regulatory capture has allows AI companies to tell the courts that their AI training consuming any and all copyrighted content is not only fair use but also required for national security, so they are effectively exempt from copyright law. The courts are corrupt, the governments are corrupt, the corporations are corrupt, we've got nothing left to stand on in the legal system.

Chances are you forfeit the right to sue them over it by signing off on their license agremeent

1000000% they are ignoring the opt out.

"We added it to the one we first assembled, right after rolling out the opt-out but before anyone had a chance to actually opt out. Obviously, you can't expect us to go through all our training sets after every opt-out to check if the model is in there and remove it, so it's not retroactive.

We won't add it to any new training sets though. Not that we'd make any: all future files are grabbed and added to the existing one in the time between the file upload hitting our content servers and the opt-out metadata reaching our file management.

So technically, it's all harvested while not opted out, we get to keep it and we can still pretend that you have a choice."

Maybe I should've studied law after all. I feel like I'd make a passable asshole.

Copyright is almost impossible to enforce as an individual over items like 3D models and art. If you create something popular it’s going to be stolen and you have to spend all your time issuing takedown notices. And that’s assuming as the other poster mentioned that you don’t surrender rights to your creation when hosting it on one of these services like CG trader.

What I don’t get is are these models not copyrighted by the creator? I’m not one to defend IP law but how exactly is it legal to not have it be opt-in?

I don't understand why you expected this to be different than all the copyrighted creations that AI companies have been sucking up for years without any permission (or opt-in) from their creators or rights holders? Why would 3D assets be any different?