Could the reason be as simple as saying that every AI model has been trained on basically the same input data sets?