"Always listening" is a fundamentally unsafe design.
Once recordings of private conversations leave the local environment and make it to the cloud, eventually they will leak. It's akin to data collection by law enforcement: once the data exists, eventually it will be abused.
Yes. Assume that all data uploaded to the cloud will eventually be compromised. That is the only safe assumption.
Not even the most responsible companies (e.g. Google) can hold out 100% of the time in the face of determined assault by government. Some of that data is going to leak to three-letter agencies, or similar.
Somewhat less responsible companies (e.g. Amazon) will leak data more often, to a wider range of threat agents.
So then we have to consider how valuable this data is. Random sampling of private conversations within the home? Sometimes innocuous -- but if the wrong moment gets leaked, the consequences are potentially life-shattering.
That seems like an extreme perspective. Kind of like the tech version of abstinence only sex ed. Sure, it's the only 100% safe way, but it's not useful for most people. It doesn't weigh the benefits against the potential cons, or even take into consideration the actually likelihood of data being leaked.
The alternative perspective you're presenting sounds to me utterly cavalier about the prospect of ruining people's lives. It's like Equifax's attitude towards identity theft: it doesn't affect their profitability, so why care?
It's because such blithe dismissal of the damage caused by data gathering is so prevalent in the industry that the likelihood of devastating compromise is so high and the costs borne by the populace are spiraling upwards.
Some data should never be collected. Some data should never even be uploaded.
The problem I have is that you are conflating potential damage and actual damage as the same thing, which is not how you accurately measure risk.
I am honestly confused as to how you interpreted my last comment as "utterly cavalier about the prospect of ruining people's lives", when all I said was that your assumption doesn't take into account the actual probability of data being leaked and it doesn't weigh any of the benefits of data collection against that risk.
Small-likelihood times many-chances times grave-consequences equals a finite but significant number of lives wrecked. A gamble you deem acceptable.
I can only hope that karma visits those who arrogate to themselves the decision to sacrifice a few of their fellow human beings: may they and their loved ones become the sacrifices.
Once recordings of private conversations leave the local environment and make it to the cloud, eventually they will leak. It's akin to data collection by law enforcement: once the data exists, eventually it will be abused.