Does anyone know if using the realtime API gpt-4o from azure or tts, llm, stt (from) azure will keep the data in our own servers ?