Because if they rule that is legal, I think the same principle should apply to humans as well.
Reading a lot is a well-known strategy to become better at writing. Thus, I argue that any reasonably skilled LLM that use the scraped text as input, but not produce verbatim copies, should not be considered as violating IP law.