Discussion: Agent Identity as a Declarative Ontological Layer

[alexanderlebed]

Dear OpenID Community,

I am writing to share a conceptual infrastructure initiative that may be relevant to your recent analytical work on identity governance for AI agents.

The OpenID document correctly identifies a structural shift: as agents move from constrained tools toward autonomous actors, traditional authentication and authorization frameworks begin to encounter conceptual limits.

My work, published at aiagentid.org, addresses a layer that sits prior to authentication and authorization.

It defines a Declarative Ontological Layer for AI Agents.

The core premise is simple:

Before we can securely authenticate or authorize an agent, we must be able to answer a more fundamental question:

What exists as an agent, and under whose declared responsibility?

The Agent ID framework proposes:
• a persistent, globally unique agent identifier
• explicit declaration of a responsible principal
• a formalized scope of delegated authority
• separation of meaning (registry) and proof (cryptographic anchoring)
• strict non-authority clause (identity ≠ certification)

This layer does not compete with OpenID standards.
It does not replace authentication protocols.
It does not attempt to solve authorization.

Instead, it attempts to formalize:
• existence
• delegation
• attribution

— before authorization logic begins.

In other words:

OpenID defines how agents authenticate.
Agent ID defines what an agent is in the first place.

As agent autonomy increases — especially with recursive delegation and cross-domain operations — this ontological grounding may become structurally necessary.

I would welcome discussion on whether such a declarative layer could complement emerging identity governance frameworks.

The project is fully public, non-commercial, and open to collaboration.

With respect,
Alexander Lebed
aiagentid.org