It wouldn't be a matter of just having twice as many phrases, because a phrase in one language could be recognised as a completely different phrase in English.
Your point also doesn't take into account mid-sentence language switches ie. simply substituting words into otherwise English sentences.
Also, your point is not valid unless the people analysing the traffic have prior knowledge of which two languages would be spoken, and therefore the system would have to cover the most common languages (at least 8 I would think). When you take into account the two points I made above, as well as trying to implement the system with more than two languages, you end up with a ridiculously large number of possibilities, as well as too many false positives (from similar sentences in other languages and mixed sentences).