I was going to guess that they accused the author of copying code from Office. Was AI used in the project? Perhaps a model regurgitated copyrighted code leading to a sternly worded notice from legal...?
Ooooh yeah. Looking through the author's past posts: "got a lot of skepticism because we're developing heavily with AI"
So AI was in use. Then the author says that following the spec alone wasn't enough to get it working, they got "active community feedback" and fed that feedback into the AI until it worked just like Word. I have to think that if there were ANY conditions under which a model might output code that Microsoft legal would threaten to sue you for, these would be them
I suspect the source code for at least some older versions of Office is absolutely in the training materials of some LLMs. There have been leaks before, and the early models were trained on the entire contents of the internet without regard to legality
I think this (if it is what happened) is a perfect demonstration of the dynamics. If you use AI to do things you couldn't have done on your own, you're copying off someone else's homework and the real risk is that you don't know who you're copying from, but they probably do.