I wonder how effective it would be just to put a bunch of data on servers meant to position the training data they’re scraping. Like, make it data that only a bot trying to get everything would find, not something that users would see or encounter.