TLDR: It’s compatible with other copy-left licenses like GPLv3. However, it’s available in multiple languages, which technically makes it more applicable.
I started using it for my own project. If you want a practical example: https://github.com/TimoKats/emmer
Yep. Either they’re ignoring the law and face no consequences or case law settles that scraping copyrighted content for LLM models is fine, in which case again, it doesn’t matter.