GenAI Observability: Keeping GenAI Honest Without Oversharing - Liudmila Molkova, Grafana Labs

This talk, led by Liudmila Molkova from Grafana Labs, explores the challenges of observability in generative AI applications. It addresses the need for capturing conversation data to ensure responsible AI behavior while balancing the issues related to sensitive information. The session provides insights on how to instrument AI applications using OpenTelemetry in accordance with the Generative AI Semantic Conventions. With a live demonstration, attendees will learn to record conversation history alongside performance telemetry and discover strategies for managing sensitive data to enhance compliance and reduce costs.