The article explains that most of these problems stem not from Google’s translation core – which compares works available in different languages – but from parts of the program where a word is missing, so that man-made online dictionaries have to be queried.
There’s an interesting question in here: does Google have editorial responsibility for their translations? They don’t claim such responsibility for web results. By emphasizing that the current change was outside of the automated translation core, Google avoided to face the question now. But as translation systems improve to higher quality levels, we might see more such issues erupt.