Spellcheck does not assert information as factual unless the author of the text being checked made that assertion already.

The same does not hold true for GPT.

Apples and oranges: The authors in question obviously missed ChatGPT generated nonsense ("Regenerate"), who's to say they didn't miss anything else? A spell checker helps you _not_ miss inconsequential mistakes. ChatGPT makes it harder to miss subtle yet consequential mistakes.

The examples here seem to go pretty far beyond spellchecking though.

I guess I understand the reasons for the disclosure requirement but if you're really just using LLMs to help you more quickly write some background text that you could otherwise do yourself, 1.) It's hard to see the harm beyond the tools that are in every word processor and 2.) I probably just wouldn't use the LLM at that point.