This is why predictions of eventual AI failure due to copyright lawsuits are likely wrong.

Assumedly, an organization training and then distributing this model cannot be stopped via copyright or breach of contract lawsuit. It may be that folks will figure out copyright-free versions of text-to-image and text-to-video, etc., models as well.

It seems that there is plenty of copyright-free data available to train a useful model. Therefore, when content creators upset about AI companies training models on their content are asked about this model, they have nothing to do but shrug.

The cat is out of the bag.