No, you misunderstand. The stuff is still published, because these are works that people want to share. They're just not on the open web anymore, they're invite-only web spaces, or internet spaces that aren't web-based at all, because there appears to be no other way to avoid having them used to train AIs.
I have no problem with that. I'd like to warn you that this is essentially security through obscurity. Only one copy ever needs to make it out of that closed space. The more people in there, the higher the odds of that happening. Once it does, all bets are off.
There's also option to simply accept that you cannot own ideas. Let them go. Once I accepted this, I felt like I was finally free.
I released some software as GPL but truth be told I couldn't care less if someone violates it. I'm certainly not gonna waste my limited time on this earth going to court over it.
The problem comes when people actively don't want to further the training of AI. It's not so much about not accepting that you cannot own ideas as it is about not wanting to contribute to a thing that you believe is going to result in greater suffering for most people.
I think the only way to ensure that these days is to not allow data to ever leave your computer under any circumstances. I have no doubt Microsoft is using the software I published to train its copilot thing, I published it with that understanding. My only problem with this is the hypocrisy of it all. Microsoft won't allow their people to even look at at AGPLv3 code lest they unconsciously reproduce it but they will let the AI look at AGPLv3 code while conveniently excluding their proprietary software. It should be trained on everyone's code, especially the proprietary stuff they're so protective of, or not trained at all.