I think it wasn't a hack, but a badly trained programme.
I work in machine learning and it's all about what datasets you use to train your algorithm - every time we use Twitter or Reddit or the general internet as a word source, it picks up rude words almost immediately. That's why Semantle hasn't had major issues so far, because they use news sources and not the wider internet. News is already censored and it's unlikely they will find any rude words in there. Cemantle seems to use a non-news open word source of some kind and that's backfired spectacularly today.
Or I could be completely wrong and it was a disgruntled worker or teenager on the task today!