Because it's impossible to identify AI-generated content, especially when you get away from visual art. (False positives exist in art but not to the scale that they do in text.)
Any "ban this content" rule would result in unpopular devs being harassed and having complaints filed that they have AI-generated content - and then they're stuck trying to prove that they actually wrote their visual novel dialogue instead of developing it through ChatGPT.
There's no point in banning something you can't reliably identify; that just makes the ban a weapon that would be used against marginalized creators.
why even allow assets that have had no significant human editing?
There's a difference between "no significant human editing" and "contains some AI-generated features" - and right now, the tag won't differentiate between those. A ban on "all AI-generated content" would include "I made 30 distinct character tokens and then used AI to create color-shifted versions of them." It would include banning randomly-generated maps in TTRPG supplements, where the map is AI-made but all the description is created by the writer.
...Would it include all images edited with Photoshop, since it has a lot of AI features now? Text edited with Google Drive's spellcheck?
I can see value in marking AI-generated assets (especially since those will be in the public domain, definitely something you want to know before adding them to another work), but banning means needing to define where the edges of "AI-generated" are - and, without a way to identify AI works, means any easy way for bullies to harass the people they don't like, by accusing them of something that can't be disproved.