Why Classic IAM Collapses for Agents: Rethinking IAM for Agentic Systems - Parul Singh, Red Hat

This talk, presented by Parul Singh from Red Hat, explores the limitations of traditional Identity and Access Management (IAM) in the context of autonomous AI agents. It highlights the need to rethink IAM models, as classic assumptions such as long-lived identities and static permissions do not apply to agentic systems that operate in dynamic environments. The speaker addresses three key questions regarding the definition of agent identity, accountability when agents act on behalf of others, and how access should adapt as agent behavior evolves. The session outlines a practical migration path towards agent-aware identity workflows and discusses insights gained from collaboration with the COSAI Secure Design for Agentic Systems working group, along with open questions for audience engagement.