This is going to be the legal challenge for the next few years: every year there are lawsuits alleging that music, drawings, etc. are too similar to someone else's work but those take time and require human judgement to decide whether something crosses the line, and the scale which tools like this offer makes that basically as feasible as it was to individually sue Napster users for sharing other people's works.

I'd expect that level of legislative response, and also would bet on lawsuits over any authorized data in their training corpus.