Copyright doesn’t protect from training on data that was scraped legally.

Which is exactly the stance I would expect from EFF, since their primary goal is an open and free internet. Scraping is a normal activity if you ascribe to that ethos.