Briefly I return, to give some corrective:
I am concerned with a new and prevalent perspective.
The idea being that it isn’t our concern,
Of how Nix is made (or used) in turn.
This is erroneous, and for reasons plain,
As the responsibility could only be ours to fulfil, or abstain.
Therefore the idea that it is not our task,
Is factually wrong, despite how policies might this mask.
To be more curt, so the message is clear,
It is our job to discuss these effects, now and here.
The dismissal otherwise, I will ignore,
As it is bad-hearted (and rhetorically a bore 
).
Just to break verse, vomit non-withstanding,
Here is a summary to complement understanding:
There has been a lot of comment about the scope and salience of Nix’s responsibility to even moderate these things, or its feasibility. I propose this is a bad framing of the issue.
First, the matter of whether something needs to be moderated should come before whether it is easy to moderate. This is hopefully obvious (not every crime can be prevented, but that does not make the act non-criminal). So let’s drop that for now, as it really has no space here.
Second, there is an idea presented above that this impinges on the openness of open source. To that I remind you the GPL also impinges on your rights to open source software, for the benefit of the project. Indeed, the intent to ban AI is based on this same logic. So away with this too! 
Thirdly, there is the idea (only approximated, not articulated) that the list I provided before is a list of particulars. In response to this, I latched on to the NixCon example before.
To my best estimation, none of my points have actually been challenged, just compared to other issues. This is not a refutation.
To see this is not a refutation we can note that, for example, Nix has ties to military companies. But the fact that Nix has ties to the military industrial complex should not give us the licence to be laissez-faire about adopting another questionable industry link. The morally conscientious approach is instead to keep stock and “not get worse”, in other words. This same rubric can be applied to all responses of this type that I have read.
Fourthly, there was some comment about how this affects access to obscuring text, and that me being against consentingly adding text to bots is in turn against good OpSec. This is not the case. Good OpSec changes dependent on the environment you are in: so again, it is a secondary concern.
Lastly, and possibly most interestingly, there was a practical conversation about how this would even be done. I had sort of neglected this thread (apologies) but I think the serious answer is that it will change over time. Several other projects are taking a hardline stance against AI commits (Gentoo and Zig come to mind, for example), and I fully expect the process to be worked out amongst the myriad communities that are all taking hardline stances against this stuff.
In particular, I think it would operate in a manner similar to what was suggested, in that its less of a “mechanistic rule” and more of “the right to reject PRs if they look AI”.
There was a response above that this would be a regression from the current policy in that it would incentivise bad actors to just not disclose that they are using AI. Which is almost true: except bad actors wont necessarily disclose with the current policy either. And so we are back to the idea that the policy is not actually practical to begin with, but is more of a statement about what the community expects.
Deepest apologies, the above para’s were vile,
I think I will be sleeping with a bucket for a while.
I hope that this summary helps contextualise my measures,
And that there is peace with no need for displeasure.
.