There are a lot of good examples, but the "Caveats" section highlights an important part. If you use LLMs frequently, you may correctly highlight 90% of posts as AI generated, but will probably falsely accuse one writer of using AI when they did not.
And how much of that is an issue to be falsely accused of using AI? Does that mean we need to use more kinder words when "accusing" something of being AI in case it isn't? Will that eventually count as harassment later?
Topics to think about.