I thought that was one of the problems with some of the art models - that you could input the right sequence of words and get exact copies of famous copyrighted images out.