Identity and Access Management for Intelligent Agents

Abstract

The rise of autonomous AI agents introduces a new class of actor in distributed systems, combining the speed and scalability of services with the autonomy and unpredictability of humans. Identity and access management models designed for users and deterministic services fail in scenarios with multiple trust domains, asynchronous operations and dynamic delegation of permissions. This paper presents GIAAI, an initiative of the EMBRAPII Competence Centre in Cybersecurity (CISSA), which evaluates traditional identity and access management tools and proposes practices, adaptations and extensions for managing agents securely.