Is it fair to characterize it as a slippery slope then?

I hadn't really considered the degradation of shared language structures due to the homogenization of outputs as a stand alone structural risk of AI use.  There is something there for sure though.  So in answer to the question 'why is it bad?' your saying that there is a degradation our ability to act collectively in conditions of uncertainty and is essentially irreversible?

Thanks for the comment!

I think probably yes. Separate to the individual level, the erosion of the collective common ground is probable as AI costs fall and 'human in the loop' becomes even more of an exception than it is now.  I guess the question then becomes, "In what ways is that bad?".
My first thought is then around the homogenization of human outputs, which could be referred to uncharitably as 'Mcdonaldisation'.  The rational offloading of cognitive labor leading to a dearth of diversity of thought, which no single person intended.

Is that useful?  I'm aware that many people in this space have thought about this much more than I have!!!