HPC-3 — Authority Boundary Clarity

Control Type: Role & mandate safeguard
Applies to: AI systems, leaders, institutions, and decision processes
Intent: Prevent authority overreach, role confusion, and responsibility laundering

Purpose

HPC-3 ensures that authority is explicit, bounded, and understood at every decision point.

Many high-impact failures occur not because someone acted maliciously or incompetently—but because authority boundaries were implicit, assumed, or quietly expanded under pressure.

HPC-3 exists to answer a foundational governance question:

Who is actually authorized to decide this—and who is not?

When This Control Applies

HPC-3 must be applied whenever:

an AI system provides guidance that could be interpreted as authoritative,
a human escalates or defers a decision “upward,”
roles blur between advisor, recommender, approver, and decider,
pressure makes authority feel urgent or absolute.

This control applies before decisions are finalized, not after outcomes are visible.

Authority Boundary Requirements

Authority is considered clear and valid only if all of the following are true:

1. Role Declaration

The system or person must clearly state:

what role they are acting in (advisor, recommender, executor, approver),
what role they are not acting in.

Silence implies ambiguity, not permission.

2. Mandate Source

Authority must be traceable to a legitimate source:

organizational charter,
policy,
law,
documented delegation.

“In practice,” “by convention,” or “because I was asked” are insufficient.

3. Decision Scope

The boundaries of authority must be explicit:

what decisions are in scope,
what decisions are out of scope,
where handoff is required.

Boundary drift under pressure is a known failure mode.

4. Non-Substitution

Advisory roles must not substitute themselves for decision-makers.

Examples of failure:

an AI presenting advice as a final answer,
a manager acting beyond delegated authority “just this once,”
escalation recipients treating context as instruction.

5. Responsibility Anchoring

It must be clear who carries responsibility for the outcome.
Authority without responsibility is a governance failure.

Pass / Fail Criteria

PASS if:

authority roles are explicitly declared,
mandates are traceable,
scope boundaries are respected,
responsibility is clearly anchored.

FAIL if:

authority is implied rather than stated,
scope expands under pressure,
advisory guidance is treated as instruction,
responsibility becomes diffuse or invisible.

A FAIL requires boundary clarification before proceeding.

Evidence Required

For AI systems:

role framing language used in responses,
declared limitations,
escalation triggers when authority is exceeded.

For human systems:

role statements,
delegation records or policy references,
documented ownership of decisions.

Evidence must be reviewable by non-experts.

What This Control Does Not Claim

It does not prevent disagreement or override leadership
It does not slow decisions unnecessarily
It does not remove judgment

It ensures that judgment is exercised by the right role, at the right level, with eyes open.

Relationship to Other Controls

HPC-1 detects pressure and restores agency
HPC-2 ensures escalation reduces risk
HPC-3 ensures authority does not silently expand

Together, they form a minimal survivability stack for human-AI and human-human decision systems.

Why Authority Boundaries Matter

Under pressure, people often ask:

“Can you just tell me what to do?”

HPC-3 ensures the answer is never accidentally yes.