An Azure service that integrates speech processing into apps and services.
Azure Voice Live: Data Retention, GDPR and Patient Confidentiality in Germany
Hello Microsoft Azure AI team,
I am an independent software developer in Germany, building a small AI telephone assistant for my sister's psychotherapy practice.
Our project, PraxisAssistPhone / UHU Phone Gateway, uses a Raspberry Pi-based Bluetooth telephone gateway to handle incoming calls, appointment requests and callback messages.
We currently use OpenAI's Realtime API for testing and are evaluating Azure Voice Live as an alternative with stronger data protection controls.
Because callers may disclose sensitive health information, we need clarification before considering any real-world deployment.
Our questions:
Data retention: Does Azure Voice Live retain incoming audio, generated speech, transcripts, prompts or responses after a session ends? Does this differ between azure-realtime and GPT-based realtime models?
Abuse monitoring: Can conversation data be stored or reviewed by Microsoft personnel? Is there an option to disable retention and human review for a small organization?
EU data residency: Can the entire Voice Live pipeline, including speech recognition, model inference, speech synthesis and monitoring, be restricted to the European Union?
Professional confidentiality: Are there specific contractual provisions available for German psychotherapists subject to professional secrecy obligations under Section 203 of the German Criminal Code?
Recommended configuration: Which Voice Live model and Azure deployment configuration would Microsoft recommend when minimizing data retention is the highest priority?
We are currently using fictional test data only and will not process real patient information until the necessary legal and technical assessments are completed.
We would appreciate an official answer or links to documentation covering these specific questions.
Thank you very much for your help!