Yes, we need not only new laws, but probably new terminology. Artists sometimes refer to their works being used in machine learning as "stealing" and "theft", which of course isn't accurate, yet there is a sense that something is being infringed upon that goes against the spirit of intellectual property that they are grasping to convey, which we perhaps need new words for.
You make plenty of valid points (I certainly agree that copyright claims can be abused), but it seems like you were avoiding mine. I asked us to consider what the reason copyright exists in the first place might be. Just now I tried googling "why do we need copyright" and this is literally the first link that came up (if you read other explanations they are similar):
https://iadt.libguides.com/copyright
Why is copyright important?
"The importance of copyright is an essential component of the modern educational experience. Copyright is important as it helps to protect the value of an author/academic/researchers work, by giving the originator of the work the ability to protect it from unlicensed or uncredited usage. This leads to the prevention of their work being copied to the degree where they cannot sell it effectively or receive credit for it. In this way, copyright fosters intellectual creativity as it provides an incentive for a creator to work freely, allowing them to gain recognition for their work as well as protecting their livelihood."
There it is. It IS about incentives. If we didn't care about preserving incentives then why would we even have copyright or ip? We wouldn't.
I still think that in the future, a clear standard for opting in or out of having a work consumed by training models is probably going to be the best way to address the interests of all. Sure, it's not going to stop the models from getting better and better. Slow them down for a short spell maybe, but it is important that we do this the right way.