Securing Azure Open AI apps in the Enterprise
summary
Practical inventory of what enterprise Azure OpenAI actually offers: Microsoft provides a Cloud Security Benchmark with 35 controls but only 7 are implemented for Azure OpenAI, encryption is on by default with an option for customer‑managed keys (via Key Vault sync patterns), inbound access uses the resource firewall and outbound egress uses an FQDN allowlist (up to 1,000 domains) managed via the REST API, request/response logging controls gate model‑training, auditing relies on microsoft.cognitiveServices policies and IaC deployment since native Terraform/Bicep coverage is limited.
AI-written summary · about
chapters (13)
- 0:00Introduction and setup
- 1:06Enterprise AI adoption challenges
- 3:50Azure Open AI hosting models
- 8:06Azure Open AI security basics
- 12:25Network security controls
- 16:18Encryption and audit logging
- 20:55Continuous auditing with Azure Policy
- 24:25Reference architecture overview
- 27:14Finding undocumented features
- 27:36Enterprise security concerns
- 29:00Vendor security landscape comparison
- 30:02CI/CD pipeline recommendations
- 31:07Q&A session
related talks
47mData Security Considerations for AI Adoption
56mEnter the Matrix: Securing Azure’s Assets
40mBuilding secure and efficient AI solutions in Azure with Partners
01h 02mWhat's New in Azure OpenAI
20mUnlocking Azure OpenAI: A 20-Minute Introduction
59mLearn Live: Introduction to Azure OpenAI Service
45mKeeping Your AI/ML Data Secure
13mAI from Microsoft's Perspective