This is not confusing. AT ALL!

The problem with your argument is that it is 100% possible to get ChatGPT to produce verbatim extracts of copyrighted works.

Exactly! This is the core of the argument The New York Times made against OpenAI. And I think they are right.