Because someone already checked, and it doesn't:

https://docs.github.com/en/github/copilot/research-recitatio...

Every literally quoted part that could infringe appears at least 10 times in the training data

This doesn't stand on its own as a defense: perhaps the 10 inputs were legitimate copies of a single source. They could be forked repos that were properly following the original's license, for example.

Or 10 different GPL projects that legitimately share code that remains copyrighted and protected by the GPL. Or 10 obscure projects that illegitimately copied code but haven't been caught.

Clearly, "10 other people did it" is no defense at all.

It might not even be "10 other people". For projects which originated outside Github, it's common for multiple users to have independently uploaded copies of the project. There's probably at least 10 users who have pushed copies of the GCC codebase to Github, for example.

> at least 10 users who have pushed copies of the GCC codebase

That is "10 other people". (Although your point stands, since there doesn't (or at least shouldn't to the point of criminal espionage) be any strong impediment preventing one person from creating 10 different accounts.)

It isn't "10+ other people wrote that", though. It's the same work, by the same person, being represented 10+ times in the training corpus.

Someone that works at Github already checked. I think asking for a more independant study is fair.