What is LLM Red Teaming? How Generative AI Safety Testing Works скачать в хорошем качестве

What is LLM Red Teaming? How Generative AI Safety Testing Works

A quick introduction to Generative AI Red Teaming (LLM Red Teaming)—what it is, how it's done, and why it matters for AI safety, security, and risk management. Learn about common jailbreaking methods, testing strategies, and model vulnerabilities. 🔗 Jailbreaking Taxonomy: https://innodata.com/llm-jailbreaking... Presented by Karen McNeil, PhD, Director of LLM Practice and Red Teaming at Innodata. Illustrations by Yevheniia Lisovaya Music by Benjamin Tissot, Bensound (License code: Y0ZTXBXFHRDIDQ4G) Innodata provides comprehensive LLM red teaming services. Visit https://innodata.com to learn more. 00:00 Introduction 00:54 Model Safety 01:53 Why is it called Red Teaming? 02:33 LLM Harms 03:34 Jailbreaking 06:04 Multimodal Red Teaming 07:57 Indirect Prompt Injection 08:57 Automated (LLM vs. LLM) Red Teaming 09:53 Why Humans Still Matter 11:30 Who Should Care About Red Teaming? 12:38 Conclusion