It is used on huge parts to harvest training data for llms. This is huge problem for smaller hosters, some open source projects, openstreetmap etc because they crawl web sites at intense depth without even basic caching which wastes lots of resources, money and often makes the service inoperable for legitimate users.
I'm shocked that this is not mentioned in the report or the comments here, in my bubble I had various reports of this problem.
You can't block those IPs because there are so many and you would also block legitimate users. That is exactly why they are using them.