ITU Workshop Hot Takes: Agentic Identity, OAuth Limits, and Human Intent

At the ITU Workshop on Agentic Identity in Geneva, industry leaders, standards contributors, and identity experts explored one of the biggest emerging questions in digital identity: How do we build trust and authorization models for autonomous AI agents? In this DIF Hot Takes session, Damian Glover and Ward Duschamps unpack key discussions from the workshop, including the growing realization that existing identity and authorization frameworks may not be sufficient for highly autonomous agentic systems. The conversation explores: • Why traditional OAuth assumptions break down in agentic environments • The shift from session-based to intent-based authorization • The concept of “human intent” as a new trust anchor • Legal and governance challenges around autonomous agents • The role of layered identity models using OAuth, workload identity, DIDs, VCs, and policy engines • Why agent identity may require “digital power of attorney” • The challenge of ephemeral agents, delegation, and revocation The discussion also highlights a growing consensus: the future of agentic identity will require layers, lawyers, and longevity. As AI systems gain autonomy, questions around accountability, authority, interoperability, and trust are becoming increasingly urgent. Topics include agentic identity, AI authorization, OAuth, decentralized identity, verifiable credentials, governance, and interoperability.