Google patents voice control that checks who is speaking before acting

1. A method implemented by one or more processors, the method comprising: receiving audio data generated by one or more microphones of a computing device, the audio data representing a spoken utterance; identifying, based on the audio data, that a guest user provided the spoken utterance; identifying, based on the audio data, a particular automation action for a particular automation device, the particular automation action corresponding to the spoken utterance; determining that the guest user is not authorized to cause performance of the particular automation action for the particular automation device; in response to determining that the guest user is not authorized to cause performance of the particular automation action for the particular automation device: refraining from initiating performance of the particular automation action; receiving additional audio data generated by the one or more microphones of the computing device, the additional audio data representing an additional spoken utterance; identifying, based on the additional audio data, that a registered user provided the additional spoken utterance; identifying, based on the additional audio data, the particular automation action for the particular automation device, the particular automation action corresponding to the additional spoken utterance; determining that the registered user is authorized to cause performance of the particular automation action for the particular automation device; and in response to determining that the registered user is authorized to cause performance of the particular automation action for the particular automation device: causing the particular automation device to perform the particular automation action.