Delay the AI Overlords - How OAuth and OpenFGA Can Keep Your AI Agents from Going Rogue

Learn how to secure AI agents and prevent unauthorized access through OAuth2 and OpenFGA authorization strategies in this 42-minute conference talk from Devoxx. Explore the critical security challenges that arise when AI agents bypass controls, abuse tool permissions, or expose sensitive data through RAG (Retrieval-Augmented Generation) pipelines. Master role-based access control (RBAC) and Fine-Grained Authorization (FGA) to enforce least privilege principles for AI toolchains, ensuring agents operate within defined boundaries. Discover how to implement credential-free tool calls using OAuth2 token exchange, allowing agents to act on behalf of users without handling raw credentials directly. Address RAG jailbreaking vulnerabilities by embedding FGA directly into retrieval workflows to prevent confidential data leakage. Understand how to establish human guardrails through asynchronous authorization workflows for auditing high-stakes AI actions. Gain practical knowledge to implement battle-tested authorization patterns that protect against real-world AI security threats, moving beyond theoretical scenarios to address today's pressing concerns about AI agent behavior and system security.