Comparison matrix
Principal vs Agent
One-sentence distinction
A principal is the entity whose authority or interests are represented; an agent acts under its own or delegated mandate.
Side-by-side matrix
| Dimension | Principal | Agent |
|---|---|---|
| Primary question | What evidence establishes the first property for a named purpose? | What separate evidence or authority establishes the second property? |
| Evidence | Purpose-specific technical, factual, or institutional records. | Independent records appropriate to the second category. |
| Authority | May be descriptive or technical and may not require legal authority. | May require a competent legal, constitutional, organizational, or operational decision-maker. |
| Currentness | Can be current, stale, disputed, unknown, or unavailable. | Must be assessed separately; the first status does not transfer. |
| Failure condition | Evidence may be authentic but incomplete or unsuitable. | Authority may exist but rely on wrong or stale facts. |
Why the distinction matters
A principal is the entity whose authority or interests are represented; an agent acts under its own or delegated mandate. Systems and institutions fail when one side is used as a shortcut for the other. The distinction determines what evidence is collected, who may decide, what can be appealed, and which failure modes must be controlled.
Common failure caused by conflation
Attributing an agent’s action to a principal without proving the mandate, scope, time, revocation state and causal path.
This error can create false confidence, unauthorized status, misattributed liability, silent loss of correction rights, or an operational claim based only on descriptive material.
Implementation consequences
- Use different fields, identifiers, and claim-status records for each side.
- Require separate evidence and currentness checks.
- Do not let a user-interface label silently merge the categories.
- Preserve correction and supersession history for both.
- Route decisions to the ecosystem authority that owns the relevant function.
Legal consequences
Current law usually attributes obligations through owners, operators, providers, principals or institutions, although allocation varies by jurisdiction and facts.
Technical evidence can inform a legal decision but cannot replace jurisdiction, legal basis, procedural authority, due process, or remedy. Conversely, a lawful decision does not make the underlying technical record accurate if the evidence is stale or defective.
Examples
- A valid signature demonstrates control over a key and payload integrity; it does not establish the truth of every signed statement.
- A registry can record a citizenship decision; the registry operator does not thereby acquire constitutional power to create citizenship.
- A static release can show that software exists; it does not prove the service is currently operating.
Sources
- Artificial Intelligence Risk Management Framework (AI RMF 1.0) — NIST; AI RMF 1.0; Published voluntary framework; revision underway. Exact claim-support entries: 2. Revalidated 2026-08-14T22:04:09Z.
- PROV-O: The PROV Ontology — W3C; W3C Recommendation 30 April 2013; W3C Recommendation. Exact claim-support entries: 2. Revalidated 2026-08-14T22:04:09Z.
Comparison claim record
Each proposition has a stable ID, status, scope, owning route, evidence relationship, currentness qualification, correction state, and synchronized JSON record. Record completeness does not make the proposition true.
Principal versus Agent
A principal is the entity whose authority or interests are represented; an agent acts under its own or delegated mandate.
Support relationship
SRC-NIST-AI-RMF· AI RMF 1.0 publication · QUALIFIES OR SUPPORTS WITHIN STATED SCOPESRC-NIST-AI-RMF· Current revision notice · QUALIFIES OR SUPPORTS WITHIN STATED SCOPESRC-W3C-PROV-O· Abstract and status · QUALIFIES OR SUPPORTS WITHIN STATED SCOPESRC-W3C-PROV-O· PROV-O at a glance · QUALIFIES OR SUPPORTS WITHIN STATED SCOPE