I expect they'll keep training on the pre-2022 corpus and supplement with new curated material. Of course they'll use AI to curate so the same process will eventually happen but probably can be slowed significantly.