The policy is sensible.

The bare minimum the author of a scientific paper must do, is to ensure that all information in their paper is well-grounded, factual, and accurate. False references generated by machine show that the author did not do the bare minimum and that, as such, their paper deserves as much trust as used toilet paper. And used TP has no place in a repository for sci papers.

Yes, even if it’s “just a refarense lol”. Plus references are an essential part of the process.

Yes, even if they “just signed it”. It’s still your paper.

Yes, regardless of this counting as “ackshyually not fraud lol” or not, like a bunch of HN users are arguing.

[And of course they’ll argue it based on “intent”, an esoteric concept mostly brought up to defend, that dead weight should get a free pass to keep causing harm. Like the user dataflow is doing there.]