Authority Continuity

The AI Agent Economy Cannot Be Trusted: What Agency Theory Teaches, and Where Runtime Authority Needs a Different Instrument
reading time: 21 minutes
The AI Agent Economy Cannot Be Trusted.
Economics offers incentives, monitoring, delegation and control. Runtime software can also make some invalid authority states rejectable at the receiving boundary.

Economics has studied agency under asymmetric information for more than fifty years. Stephen Ross gave an early formal treatment of the principal-agent problem in 1973. Jensen and Meckling’s 1976 paper defined agency costs and integrated agency theory with property rights and finance to analyze ownership structure. Holmström, Grossman, Hart, Moore, Tirole and many others subsequently developed major parts of contract and organization theory. Work in this broader tradition later received Nobel recognition, including the 2016 prize to Oliver Hart and Bengt Holmström for contributions to contract theory.

Designing PIC-X: Exposing Configuration through .well-known/pic-x-configuration
reading time: 8 minutes
Designing PIC-X: Exposing Configuration through .well-known/pic-x-configuration.
Designing PIC-X. Exposing Configuration through .well-known/pic-x-configuration.

PIC-X exposes discovery at two levels: a server control-plane document and a per-realm PIC-X discovery document.

The server is not an issuer. A realm is an isolated trust domain and issuer boundary. Realm discovery publishes issuer-scoped endpoints, keys, token-exchange metadata, and PIC authority and continuity capabilities.

/.well-known/server-configuration
        |
        +--> realm: acme
                |
                v
/realms/acme/.well-known/pic-x-configuration

Discovery Model

The server-level document describes the PIC-X instance. The realm-level document describes the issuer clients use for PIC exchange and verification.

Designing PIC-X: Deriving an Initial PIC Context of Authority
reading time: 14 minutes
Designing PIC-X: Deriving an Initial PIC Context of Authority.
Designing PIC-X. Deriving an Initial PIC Context of Authority.

PIC-X receives an OAuth access token at the selected realm token endpoint, validates it, and derives the initial PIC Context of Authority, or PCA. A PCA is the logical Context of Authority. Its signed representation is a PIC PCA COSE. PIC-X then returns a realm-signed PIC Token JWT carrying a settled PIC Continuity COSE in pic.root.

Designing PIC-X: From Specification to Architecture to Code
reading time: 3 minutes
Designing PIC-X: From Specification to Architecture to Code.
Designing PIC-X. From Specification to Architecture to Code.

PIC-X stands for Provenance Identity Continuity Exchange.

Verifiable Authority Continuity across execution boundaries.

PIC has reached a point where the questions are becoming practical:

How can we use it? How can we test it? What should a real implementation look like?

Answering those questions requires more than specifications. It requires building the components that enable software engineers and architects to experiment with PIC in real systems.

From Hardy’s Compiler to Hardy’s AI Agent: When the Threat Model Changes, the Confused Deputy Returns
reading time: 87 minutes
From Hardy’s Compiler to Hardy’s AI Agent.
From Hardy’s Compiler to Hardy’s AI Agent.

The question. This article examines one narrowly defined security property: whether authority propagated across time and multiple execution boundaries remains verifiably attributable to the concrete execution and request for which it was granted.

The change of question in authority propagation: from actor selection to execution continuity

One scope condition should be kept in mind throughout this article: the following discussion of who concerns only the propagation of authority after initial authentication and authority origination. It addresses what happens when already-established authority is handed off, exchanged, resumed, attenuated, or otherwise continued across a later execution boundary. It does not claim that identity is unnecessary for authentication, initial authorisation, eligibility, accountability, revocation, or other security properties.

Trusting AI Agents, Who Is Acting?
reading time: 10 minutes
Sample Work Pool
Who Is Acting?

AI agent security is usually approached as an identity problem: the question becomes “what is the agent’s identity, and what is it allowed to do?”. This framing is inherited from decades of human-centric and client-centric authorization design, and it works reasonably well as long as the entity being authorized is stable, persistent, and accountable in its own right. AI agents are none of those things in practice, and the framing produces a steady accumulation of edge cases that identity-centric security models keep trying to patch without changing the underlying assumption.