Episode 502 - Azure Open AI and Security
Azure Open AI is widely used in industry but there are number of security aspects that must be taken into account when using the technology. Luckily for us, Audrey Long, a Software Engineer at Microsoft, security expert and renowned conference speaker, gives us insights into securing LLMs and provides various tips, tricks and tools to help developers use these models safely in their applications.
Media file: https://azpodcast.blob.core.windows.net/episodes/Episode502.mp3
YouTube: https://youtu.be/64Achcz97PI
Resources:
AI Tooling:
- Azure AI Tooling Announcing new tools in Azure AI to help you build more secure and trustworthy generative AI applications | Microsoft Azure Blog
- Prompt Shields to detect and block prompt injection attacks, including a new model for identifying indirect prompt attacks before they impact your model, coming soon and now available in preview in Azure AI Content Safety.
- Groundedness detection to detect “hallucinations” in model outputs, coming soon.
- Safety system messagesto steer your model’s behavior toward safe, responsible outputs, coming soon.
- Safety evaluations to assess an application’s vulnerability to jailbreak attacks and to generating content risks, now available in preview.
- Risk and safety monitoring to understand what model inputs, outputs, and end users are triggering content filters to inform mitigations, coming soon, and now available in preview in Azure OpenAI Service.
- AI Defender for Cloud
- AI Red Teaming Tool
AI Development Considerations:
- AI Assessment from Microsoft
- Microsoft Responsible AI Processes
- Define Use Case and Model Architecture
- Content Filtering System
- How to use content filters (preview) with Azure OpenAI Service - Azure OpenAI | Microsoft Learn
- Azure OpenAI Service includes a content filtering system that works alongside core models, including DALL-E image generation models. This system uses an ensemble of classification models to detect and prevent harmful content in both input prompts and output completions
- The filtering system covers four main categories: hate, sexual, violence, and self-harm
- Each category is assessed at four severity levels: safe, low, medium, and high
- Additional classifiers are available for detecting jailbreak risks and known content for text and code. JailBreaking Content Filters
- Red Teaming the LLM
- Create a Threat Model with OWASP Top 10
Other updates:
- Los Angeles Azure Extended Zones
- Carbon Optimization
- App Config Ref GA
- OS SKU In-Place Migration for AKS
- Operator CRD Support with Azure Monitor Managed Service
- Azure API Center Visual Studio Code Extension Pre-release
- Azure API Management WordPress Plugin
- Announcing a New OpenAI Feature for Developers on Azure
Published on:
Learn moreRelated posts
Azure Migrate Execute
From Manual Testing to AI-Generated Automation: Our Azure DevOps MCP + Playwright Success Story
In today’s fast-paced software development cycles, manual testing often becomes a significant bottleneck. Our team was facing a growing backlo...
Cognitive services and Azure ML for Dataflows will be fully retired by September 15th, 2025
This blog is outlining the depreciation announcement for Azure ML and Cognitive services using dataflows.
Azure Developer CLI: From Dev to Prod with One Click
This post walks through how to implement a “build once, deploy everywhere” pattern using Azure Developer CLI (azd) that provisions...
Azure Migrate assessments
AI Builder – Invoice processing and Invoices document type to begin using Azure
Starting on July 21, 2025, the prebuilt model invoice processing and invoices document type (built on Azure Document Intelligence 4.0) will be...
Dataverse: Learn How to Implement Azure Durable Functions – Payment Scenario
Azure Durable Functions is an extension of Azure Functions that offers specialized capabilities, including statefulness, orchestration, handli...
Build reliable Go applications: Configuring Azure Cosmos DB Go SDK for real-world scenarios
When building applications that interact with databases, developers frequently encounter scenarios where default SDK configurations don’...