What content was used to train the Video AI models?

From what I know, none of that matters since the output of TVAI, is the input. There is zero percent chance that TVAI will start to output copyrighted works that it was trained on.