Ensuring LLM Safety | Lunchtime BABLing 60

🎯 FREE AI Training Assessment — Find Your Best Certification Path ➡ https://shea-1mb3pmep.scoreapp.com/?utm_source=youtube&utm_medium=video&utm_campaign=ai_readiness_check 🎙️ Ensuring LLM Safety | Lunchtime BABLing 60 With growing pressure, organizations developing or deploying Large Language Models (LLM)-powered systems face increasing demands. In this Lunchtime BABLing episode, BABL AI CEO Dr. Shea Brown explores one of the most urgent issues in AI governance—how to ensure the safety of LLMs. 📊 New to BABL AI? Start with the FREE AI Training Assessment: https://shea-1mb3pmep.scoreapp.com/ 👉 Lunchtime BABLing listeners can save 20% on all BABL AI courses with code BABLING20 📚 Courses: https://babl.ai/courses/ 🌐 Visit BABL AI: https://babl.ai/ 📩 Subscribe to The Algorithmic Bias Lab mailing list: https://www.algorithmicbiaslab.com/ 🔗 Follow BABL AI: https://linktr.ee/babl.ai ⏱️ Chapters 00:00 – Intro: Why LLM Evaluations Matter for Risk & Compliance 00:55 – Overview: Ethics, Risk & Regulatory Pressures (EU AI Act, Colorado, NY) 01:32 – Key Takeaways: Evaluations, Sociotechnical Mindset & Documentation 02:31 – Why You Can’t Wait for Standards—Focus on First Principles 04:02 – Regulatory Pressure: EU AI Act Article 9 & NX3 Obligations 05:30 – Why Evaluations Are Essential for AI Systems 07:33 – A Basic Framework for LLM Testing & Documentation 08:30 – From an Auditor’s Perspective: What You Need to Prove 09:18 – What to Document: Context, Users, Use Cases & Fail States 10:19 – Introducing the CIDA Narrative: Context → Input → Decision → Action 12:19 – How to Run a Risk Assessment for LLMs 13:07 – Common LLM Risks: Confabulations, Toxicity & Robustness 15:51 – Grouping Risks: Using NIST & Custom Categories 18:14 – Using HELM Benchmarks and When to Customize Tests 19:30 – Prompt/Response Testing & Quantifying Performance 21:44 – Test Coverage Strategy: Focus on Both Risk & Performance 22:24 – Parameter Space Thinking: Mapping Real-World Complexity 24:28 – How to Probe the AI System’s Full Behavioral Landscape 26:08 – Capturing the Full Chain: From Inputs to Consequences 27:42 – Outro: Subscribe for More on AI Risk, Governance & Testing 📌 What You’ll Learn 🔵Why LLM evaluations are critical for compliance and ethics 🔵How to document context, users, and failure points effectively 🔵The role of socio-technical thinking in AI risk assessments 🔵How to apply the CIDA framework for robust AI evaluations 💬 Join the Conversation 👍 Like this video if you want more expert insights into AI risk, safety, and compliance 🔔 Subscribe to Lunchtime BABLing for weekly discussions on AI governance and regulation 💬 Share your thoughts in the comments—how should regulators balance LLM innovation with safety? 🔖 Keywords LLM safety, Large Language Models, AI compliance, AI risk assessment, EU AI Act, AI governance, CIDA framework, AI evaluation, AI regulation, AI auditing, Shea Brown, BABL AI #AI #ResponsibleAI #AIAudit #lunchtimebabling