The law will consider "intent". By your logic, web browsers are infringing.

Can you use curl to infringe on copyright? Yes. Is every time you use curl copyright infringement? No. Can you in theory tell when you are infringing with curl? Yes.

Can you use copilot to infringe? Yes. Is every time you use copilot copyright infringement? No. Can you in theory tell when you are infringing with copilot? *No*

my response above about intent is wrong, but the rest still holds imo.

So the difference is entirely in being able to find out whether you're infringing or not?

Surely that's solvable with a EULA that passes the responsibility onto the user to search?

I'm not a lawyer, but if you provide a platform that enables infringement that's different than if you provide a tool that could enable infringement.

Popcorn time vs. bittorrent.

And you are right the EULA could say "it's up to the end user to confirm you can use this code". But then how do you verify? That slows down "productivity" where copilot promises "speeding up" productivity.

Yep, makes sense! I guess we'll see what arguments the court finds convincing. I, for one, hope Copilot stays, but if we can delay its destruction long enough I think we'll get open-source models that will give us this for free. And then the cat will be out of the bag.

https://github.com/moyix/fauxpilot is working on an open source variant. Based on https://github.com/salesforce/CodeGen