Training on open-source code isn't a problem in and of itself. As I said before, the training is likely transformative. Co-pilot operates as little more than a statistical recommendation engine.
The biggest issue is the final output. I don't know if Microsoft did anything in particular to solve the issue of verbatim outputs. But whatever the case, I'm of the view that since Co-pilot is now production-ready, Microsoft may now be held legally responsible should such problems arise.
The biggest issue is the final output. I don't know if Microsoft did anything in particular to solve the issue of verbatim outputs. But whatever the case, I'm of the view that since Co-pilot is now production-ready, Microsoft may now be held legally responsible should such problems arise.