Agent
A software system that takes consequential action toward a goal without a human approving each action. Used in this corpus only where the qualified form would be imprecise; every governance claim about an agent attaches to a qualified term instead.
Why it matters
Bare 'agent' is the most overloaded word in this field and this entry exists to say so rather than to settle it. It is used three incompatible ways at once: as a legal term of art, as a software architecture, and as a marketing label for a model with tools. The legal sense is the dangerous one. Restatement (Third) of Agency §1.01 defines agency as a fiduciary relationship between persons, and current law does not recognize a model as a legal agent — so calling a system an agent does not make it one, and an instrument drafted as though it did rests on a premise the law does not support. Qualify the term.
This term is a QIS Canonical Extension. It does not currently appear in the Model Context Protocol specification, the A2A specification, the AIEOG Shared AI Lexicon, or any instrument in our standards register. We are proposing it because we have documented the gap it fills, and we label it as ours everywhere it appears. Adopt it, argue with it, or ignore it — but do not cite it as industry consensus, because it is not.
What it is not
These are routinely confused with Agent. The distinctions are not pedantic — each one has consequences for how a system is governed.
Agency law describes a fiduciary relationship between persons. No model is a legal agent, and nothing in this corpus asserts otherwise. This is the conflation that does actual damage, because it implies responsibility traveled with the delegation.
A model is a function from inputs to outputs. An agent is a system that acts on those outputs, holds state, and commits.
This term names the category. Agent Identity names the property that a specific one is distinguishable and attributable — which is the precondition for governing it at all.
An assistant proposes and a person disposes. The governance question arises exactly where that stops being true.
Relationships
Typed edges into the rest of the ontology. These are what make the canon traversable rather than merely readable.
| Verb | Target | Meaning |
|---|---|---|
addressedBy | Identity | Inverse of addresses. The subject speaks to the object as a question or concern. |
relatedTo | Agent Identity | An association too weak or too general for a stronger verb. |
governedBy | Authority Grant | The subject operates under the authority or constraint of the object. |
Record
| Canonical identifier | QIS-TERM-00060 |
| Status | QIS Canonical Extension |
| Adoption | Contested |
| Domain · Layer | Identity & discovery · Governance |
| Origin | QIS Canonical Extension. Entered to record a position the canon was already taking silently: every other entry qualifies the word (Agent Card, Agent Identity, Agent Registry, Agent Memory) and none used it bare, which left the most-conflated word in this field with no entry at all. |
| Semantic aliases | None recorded. |
| First published | 2026-08-02 |
| Last reviewed | 2026-08-02 · 180-day cycle |
Cite the identifier, not the URL. Identifiers are stable; URLs may change. This entry is free to read, quote and index under the dual license. Corrections to editor@qisfund.com are published.