Comment by Findecanor

In legal terms, that is not entirely correct. In practice, it is however. For now.

EU (which is not the whole of Europe) has regulation that allows a copyright owner to opt out of data mining for AI training. But the framework is incomplete: there does not exist a generally agree-upon method to actually opt out. There are a few protocols and file formats from a couple of organisation but none which has been given any official status. While a publisher may use one, a web scraper might support only another.

Japan has traditionally been quite strict on copyright law. I would not be surprised if the law would get tightened to explicitly disallow AI training on copyrighted works.