Can AI use bad language without getting in trouble?