And, by establishing legal precedent that AIs can’t be trained on copyrighted content without purchasing licenses as if the content were going to be redistributed, we’ve ensured that people who aren’t backed by millions of dollars won’t be able to build their own AIs.