I think we can already see the early innings of this with large API providers figuring out how to calibrate post-training techniques (RHLF, constitutional AI) between economic usefulness and the "mean" of western morals. Tough to go against economic incentives