AI model does not know special insight into how model itself was trained. All it tells you is it's prediction of expected explanation.