An Azure service that provides access to OpenAI’s GPT-3 models with enterprise capabilities.
Azure OpenAI GPT-Realtime 1.5: Clarification on Data Retention and Privacy Policies
Problem description
I am seeking clarification on how Azure OpenAI's GPT-Realtime 1.5 deployment handles data, specifically regarding the retention and storage of audio inputs, outputs, transcriptions, prompts, and generated responses. I need detailed information on whether these data types are stored after a session ends, the duration of retention, and how abuse monitoring and content logging are managed.
Environment
Azure OpenAI Service; Model: gpt-realtime-1.5; Usage: Realtime API / audio conversation; Deployment: Global Standard; Region: Not specified.
What I've already tried
I reviewed Microsoft's general GDPR/privacy guidance for Cognitive Services and specific Azure OpenAI data handling policies. I examined available documentation indicating default retention periods of up to 30 days for prompts and completions, but found no explicit details for audio inputs, outputs, or transcriptions for gpt-realtime-1.5. I also checked the support responses and product documentation confirming support for global deployment but not addressing all privacy-specific questions.
Current status
I am awaiting a definitive, official statement from Microsoft that clearly details how audio and conversation data are handled, retained, or deleted in the context of GPT-Realtime 1.5. I need this information to ensure compliance with GDPR and to inform my privacy documentation.