Protected group bias and stereotypes in Large Language Models