Weakness of Current AI image generators

Recently, I noticed some AI image generators tend to perform badly when it comes to items that are not closely related to humans. I generated these two images using gencraft.com and you can easily see that while the women and their clothing look ok, the errors at the rifle and sword are significant. It looks like AI didn't know much about the basic structure & function of a rifle / sword. Is it because the developer didn't put enough training resources related to weapons into the system, or is it because the system automatically pick an important topic (for instance, here the important topic is the women and their clothing) and spend less effort on other items (rifle & sword)?