IAM, Agent: Identity for Autonomous AI - Matthew Bates, Cofide

Abstract

First there were chatbots, then LLMs and now we’re beginning to hear everyone talk about “agents”, where multiple AI agents collaborate and execute tasks autonomously. As AI systems evolve toward multi-agent architectures, robust identity and access management (IAM) becomes critical for security. While these share similarities with microservices, AI agents introduce unique challenges around dynamic capabilities, trust and the interplay between human and agent identities. This talk explores applying zero trust principles to AI agent workloads using CNCF projects like SPIFFE/SPIRE and emerging IETF standards (WIMSE). We’ll explore dynamic identity provisioning, agent-to-agent authentication, and cryptographic attestation. Through hands-on demonstrations, you’ll learn how to implement secure, standards-compliant identity management in your multi-agent AI systems, addressing both familiar distributed systems challenges and novel security considerations.