Unfortunately, the trend is to stick a camera and microphone (and compute/network stack) on every consumer device sold to you, including TVs and refrigerators. Not sure what the solution is, since the vendors can make a plausible argument for the convenience provided by voice control and additional features enabled by object detection. Perhaps some combination of local (rather than cloud) inferencing, immediate deletion of data, and privacy laws that clearly assign ownership of data to the consumer and dramatically limit what telemetry data can be sent back?