It's like saying that someone buys a car that's powered by an explosive liquid, and then complains when all the gas in the tank explodes at the same time.
Given that, it would be ridiculous to expect paying customers to switch it off when not in use. The onus on Amazon is to protect their customers' privacy while the device is being used as intended.
If you really want to pick apart the car analogy, maybe this is closer to having automatic start and driving capability on a car, but the car sometimes randomly starts and drives away. Are you really going to blame the customer for that, even though they could disable the feature every time they exit the car?
I think for some, at least users like me would like a feature similar to "su". I'd like to confirm that Alexa should start doing stuff if it has not been confirmed for x number of minutes. I think the real problem with these home assistant devices is they are not designed with a confirmation message. This would be annoying in some cases, but surely a balance can be struck.
I'm not trying to be reductive at all about the car analogy. When you turn on a car, you acknowledge that the engine is running. You can also mute an echo if you have the foreknowledge that you will not be doing activities that require voice activation.
But there is a hilarity of, on one hand being privacy centric, and with the other hand, planting a listening device in ones home.