There is no on-device speech recognition. All of it is done through Siri. That's why I'm looking to hear from someone from Apple on how that data is stored/used.