I've got a Moto-X android phone and I changed its voice activation function to be the phrase "Kirk to Enterprise" (geeky yes, but also very unlikely to be spoken in public) and what I've found is that the limited ML classifier in the phone hits it regularly. I suppose I could try something in Klingon. In the referenced article, the unit activated unintentionally as well.
That is a nice way to say that your argument is unconvincing in that it depends on the ML classifier to recognize the 'wake' phrase reliably, and I have ample evidence that it has a number of false positives. And when it "wakes up" it is trying to figure out what you're asking it to do, and if you're not then you are perhaps in danger of it doing something you didn't intend. I eventually turned this feature off on my phone entirely, I could not make it reliable.
And I expect to hear interesting stories of that time when someone's phone called or did something unexpected, because the technology makes it inevitable.