I think it’s “legally grey” in the sense that governments have largely made no policies one way or the other on the data harvesting. It’s not banned, but it’s not openly encouraged either, and there’s no real legal precedent to point to for this specific matter besides the general data harvesting big tech does.

The area with the largest similarity I feel is music sampling, and as far as I know, the music industry was very quick to ensure that data harvesting for AI had to follow the same copyright laws as sampling.