> I'm pretty sure now that the debate of AI ownership is a foregone conclusion - nobody owns AI outputs.

but at the same time, they put ToS that you may not train a new LLM using the output of their LLM...