Many artificial intelligence tools were trained on freely-available digital content. That might be legal, but is it ethical?

Where’s the line between fair use and commercial exploitation when it comes to scraping the web to train artificial intelligence models?