“JPMorgan's proprietary internal dataset is 150 petabytes, which is significantly larger than the less than 1 petabyte internet dataset used to train GPT-4.”