> But training an AI model on media (code or otherwise) is not copyright infringement, so the license is irrelevant.
Well, maybe. But even if we assume that this is true, when anyone later uses the AI to reproduce a copy of the code, a copy has been made and copyright has been infringed.
If I need a code to loop over 10 lines, I'll code a for loop the same way regardless of what I'm developing.
Define for me, at what point of complexity, does code gets Copyrighted?
The things copilot is outputting is literally small chunks of code that needs a lot of cleanup afterwards. Is not like I type "Build twitter for me" and BAM, I got a working clone of twitter.
Every line that you produce is copyrighted by you, automatically. There is no "obviousness" test for copyright: if you decide to code a for loop, then that for loop is copyrighted the moment you type it in.
copilot is outputting literally small chunks of code that needs a lot of cleanup afterwards
If you start from copyrighted chunks, then clean them, you're still violating people's copyright. Multi-million dollar lawsuits have been fought over people using small samples from other people's music, cleaning them, and releasing them as parts of their own song.
Well, maybe. But even if we assume that this is true, when anyone later uses the AI to reproduce a copy of the code, a copy has been made and copyright has been infringed.