# Authority–Identity Separation **Domain:** Identity / Delegation / Rights **Doc Type:** Canonical Concept Node **Maturity:** Evolving **Related:** [[Representational Authority]], [[Descent vs Derivation]], [[Reconstructed Person]], [[Citizenship]], [[Rights]] --- ## Definition **Authority–Identity Separation** is the rule that what an entity is, what it may do and whom it may represent are different questions. Identity does not automatically confer authority; delegated authority does not transform the agent into the delegating person; moral standing does not create competence or political jurisdiction. ## Separable Axes A system may have: - identity without authority to bind another person; - delegated authority without consciousness or personhood; - personhood without competence for a particular office; - lineage without permission to impersonate its source; or - technical privilege without legitimate representation. The Machine choosing Root's voice in [[Person of Interest S5E10 — The Day the World Went Away]] exemplifies relational inheritance without identity equivalence. [[Halores]] exemplifies lineage and social impersonation without continuity as Charlotte Hale. A reconstructed person may deserve protection while possessing only narrowly granted authority—or none—to act as the historical source. ## Governance Interface NIST's work on agent identity and authorization independently reinforces the operational need to distinguish authentication, permission and action on behalf of a principal. California's digital-replica law separately protects specified posthumous voice and likeness uses without declaring the replica to be the deceased person. An August 14, 2026 individual Internet-Draft, _Architectural Requirements for Supporting AI Agents on the Internet_, makes the separation unusually explicit: > “Identity, authentication, authorization, delegation, capability, intent, trust, payment, and audit are distinct concepts and MUST NOT be implicitly conflated.” Its `REQ-AUTH-1` separately requires an architecture to distinguish an agent's identity from the human or organization for whom it acts. It also treats intent as insufficient authorization. This is a direct protocol-level analogue of the concept: an agent can be independently identifiable while exercising authority that originates elsewhere. The evidentiary status must remain visible. This is an **active individual Internet-Draft**, not an adopted standard, IETF consensus document or IETF-endorsed position. Its value here is as evidence that Internet architecture work is independently encountering the same distinction. ## Key Insight **Who an entity is does not settle whom it may bind, and permission to act does not settle who the actor is.** ## See Also [[Representational Authority]], [[Functional Privilege]], [[Proof of Descent]], [[Charlotte Hale]], [[Halores]] ## Sources / Provenance - [NIST AI Agent Standards Initiative](https://www.nist.gov/artificial-intelligence/ai-agent-standards-initiative). [[wiki/Westworld S3E6 — The Model Beneath the Model|Decoherence]] provides a four-layer demonstration: [[wiki/William (Westworld)|William]] is the continuing subject, the staged [[wiki/James Delos|James Delos]] is a reconstructed evaluator, the AR system is the interaction environment, and the institution is the administrator. Technical authority over the environment does not make the institution the author of William's identity, and resemblance to Delos does not establish Delos's continuity. - [California AB 1836](https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB1836). - [IETF Datatracker: Architectural Requirements for Supporting AI Agents on the Internet](https://datatracker.ietf.org/doc/html/draft-daniel-ai-agent-internet-architecture-00) — individual Internet-Draft, work in progress. - [IETF Datatracker: AI Agent Authentication and Authorization](https://datatracker.ietf.org/doc/draft-klrc-aiagent-auth/) — individual Internet-Draft, work in progress.