> how do they differ from a software engineer who copies and pastes code?

They differ because author of the code did not agreed to use his code to teach LLM.

> Second, if the code is hosted anywhere else, there is no guarantee that Copilot (or another model) won't learn from that

This issue should be resolved not by author of the code, but by Copilot or any other LLM team.

> Third, for me, the crucial part of open-source code is maintenance.

Even if author is against it? This is a valid argument, but results is not very different from pirating of software.