Something they should be able to do now, but do not seem to, is to allow you to train Siri to recognize exactly your voice and accent. Which is to say, to take the speech-to-text model that is listening and putting it into the Siri integration API, to both be 99.99% accurate for your speech and to recognize you and only you when it comes to invoking voice commands.
It could, if it chose to, continue to recognize all voices but at the same time limit the things the non-owner could ask for based on owner preferences.
This is really easy to do: it's just an embedding of your voice. So typically like 10/30 sec max of your voice to configure this. You already do a similar setup for faceId. I agree with you, I don't understand why they don't do it.
It could, if it chose to, continue to recognize all voices but at the same time limit the things the non-owner could ask for based on owner preferences.