What service/engine/server does the Web Speech API use for voice recognition?

Speech recogition is accessed via the SpeechRecognition interface, which provides the ability to recognize voice context from an audio input (normally via the device's default speech recognition service) and respond appropriately.

Where does the Web Speech API send the audio input? Does it recognize the audio locally or it uses a Google server?

Android has an in-built speech recognition engine/service, can Web Speech API use that?