It does however come with this risk, that they let data, or Jupyter notebook output or etc. get committed.
I'm not saying what happened here is excusable -- it's not, it's ridiculously bad. But I can see a few ways how it'd happen.
To me the equal or better question is why does this person have PII in the first place? It's not necessary to do analysis. Someone should have masked or removed it before this person ever got their hands on it. There's no way a PII dataset is needed for training.