Your AI Agents Are Overprivileged - The Case for Fine-Grained Authorization

Learn how to implement Fine-Grained Authorization (FGA) to secure AI agents and prevent data breaches in this 55-minute conference talk from NDC Copenhagen. Explore the critical security vulnerabilities that arise when AI agents have excessive access privileges and discover why traditional access control methods fail to handle contextual, document-level permissions at enterprise scale. Dive into practical implementations using OpenFGA and LangChain to create robust security boundaries for RAG (Retrieval-Augmented Generation) and agentic AI systems. Master techniques for multi-tenant isolation, data leakage prevention, and comprehensive auditability while maintaining performance across billions of access decisions. Examine real-world scenarios for both internal copilots and customer-facing AI applications, understanding how to balance AI functionality with strict data governance requirements. Gain insights into scaling authorization systems that can handle the complex, contextual nature of modern AI workloads without compromising on security or performance.