Explore the critical challenges of authentication and authorization for AI agents in this 14-minute conference talk from the AI Engineer World's Fair. Learn how traditional access control patterns like OAuth, session tokens, and role-based access fall short when AI agents are calling APIs, submitting forms, and sending emails on behalf of users or organizations. Discover how machine identity is evolving to address the unique security requirements of agent-powered systems, including strategies for thinking about authentication for non-human actors and authorizing actions when the decision-maker is an LLM or headless service. Gain practical insights from WorkOS and Cloudflare experts on assigning, managing, and revoking agent identity and access in real-world scenarios. Walk away with actionable tools and mental frameworks to build secure, auditable, and scalable systems that can safely handle AI agents operating with appropriate permissions and constraints.