Is your LLM-powered app safe? Evaluate it! | DEM522

From the creator

Join us for a live demo where we'll show you how to add safety evaluations to apps built on LLMs! Using the powerful Azure AI Evaluation SDK, we'll demonstrate how to make sure your app responses are safe and free from harmful content. We'll walk you through the process step-by-step—from setting up your Azure AI Project to simulating app responses with the AdversarialSimulator, and using the ContentSafetyEvaluator to catch any issues. To learn more, please check out these resources: * https://aka.ms/build25/plan/ADAI_DevsAdvPlan 𝗦𝗽𝗲𝗮𝗸𝗲𝗿𝘀: * Pamela Fox 𝗦𝗲𝘀𝘀𝗶𝗼𝗻 𝗜𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻: This is one of many sessions from the Microsoft Build 2025 event. View even more sessions on-demand and learn about Microsoft Build at https://build.microsoft.com DEM522 | English (US) | AI, Copilot & Agents #MSBuild Chapters: 0:00 - Discussion on LM Powered Apps 00:00:56 - A Basic Application Demonstration on Data Retrieval 00:01:28 - Exploring Content Filters and Their Efficacy 00:04:23 - Introduction of Automated Red Teaming by Azure AI 00:05:36 - Configuration of Red Team Class Parameters 00:07:27 - Moderate Complexity Transformation Example 00:09:33 - Safety Features of GP4-O Mini Model on Azure 00:13:21 - Exploring Content Safety with URL Escaping Attack Results 00:13:50 - Documentations and QR Code for Further Learning

Published 2025-05-23 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.