It really isn't that bonkers. A lot software thought is about licensing. See GPL and Creative Commons and all that stuff thats all about how things can be profited from/responsibilities around it. Benefiting from free data is one thing. Privately profiting at the expense or not sharing the capability/advances that came from it is another. Willing to bet there's GPL violations via the training sets.

Is it even possible to attach licenses to text posts on social media?