Unless these devices have enough performance to do on-device speech recognition, they could have to stream audio on an ongoing basis. I think the data stream f

Unless these devices have enough performance to do on-device speech recognition, they could have to stream audio on an ongoing basis. I think the data stream from uploading compressed audio for an extended period would be difficult to hide.