Massaging AI language models for fun, profit and ethics

Do AI language models really demonstrate intelligence? What about morality? Is it ok to tweak them, and if yes, who gets to do this, and how do the rest of us know?