Human presence is not the same as admissible human governance.

Organizations frequently describe systems as having human oversight without specifying what that actually means.

One system may simply record that a person viewed an output.

Another may require an authorized professional to review preserved evidence before execution.

Another may ensure that human authority, evidence, refusal, escalation, commitment, execution, and outcome are all connected through a replayable governance record.

All three may be described publicly as:

“Human-in-the-loop.”

They are not equivalent.

The TA-14 Admissible Human Intervention Architecture therefore establishes six Human Intervention Admissibility Levels: AHI-0 through AHI-5.

These levels describe progressively stronger evidence that human participation is actually capable of governing consequence.

AHI-0 — Presence Only

AHI-1 — Recorded Human Review

AHI-2 — Bounded Human Intervention

AHI-3 — Admissible Human Intervention

AHI-4 — Pressure-Survivable Human Intervention

AHI-5 — Route-Complete Human Governance

The purpose is not to rank humans.

It is to classify the governance condition surrounding human intervention.

Why Levels Are Necessary

The statement:

“A human reviews every decision.”

does not tell us:

what the human sees,

whether the evidence is current,

whether meaningful verification is possible,

whether the human is competent,

whether authority exists,

whether scope is bounded,

whether refusal is possible,

whether escalation works,

whether execution can actually be stopped,

whether the human determination binds,

or whether the outcome is preserved.

AHI levels make those distinctions visible.

AHI-0

Presence Only

At AHI-0, a human exists somewhere in or around the route.

Examples include:

a person monitoring a dashboard,

an operator receiving notifications,

a manager being copied on reports,

a reviewer theoretically available,

or a person nominally designated as responsible.

There may be no evidence that the human:

meaningfully reviewed the condition,

possessed standing,

could verify the information,

could influence execution,

or could prevent consequence.

AHI-0 establishes:

Human presence.

It does not establish:

Human governance.

AHI-0 PRINCIPLE

A person being nearby does not make a consequence admissible.

AHI-1

Recorded Human Review

At AHI-1, the system can establish that a human actually interacted with the decision route.

The record may preserve:

identity,

timestamp,

review event,

acknowledgment,

comment,

approval,

or rejection.

This is stronger than AHI-0 because human participation is attributable.

But serious questions remain.

Did the person see sufficient evidence?

Was the evidence current?

Did the person understand the decision?

Did they possess authority?

Could they refuse?

Could they stop execution?

A timestamped approval proves that an approval event occurred.

It does not prove that the approval was admissible.

AHI-1 PRINCIPLE

Recorded participation is evidence of participation — not evidence of sufficient governance.

AHI-2

Bounded Human Intervention

At AHI-2, the human's role begins to become structurally governed.

The architecture identifies:

the human role,

the decision class,

the person's authority,

the permitted scope,

and the intervention options.

The system can distinguish whether the human is acting as:

observer,

reviewer,

diagnoser,

determiner,

authorizer,

intervener,

escalation authority,

or consequence authority.

The human may have explicit states such as:

ALLOW

HOLD

DENY

ESCALATE

rather than a generic approval button.

This is a substantial improvement.

But AHI-2 does not necessarily prove that the evidence supporting the human's determination is itself sufficient or continuous.

AHI-2 PRINCIPLE

The human role is bounded, but admissibility of the intervention is not yet fully established.

AHI-3

Admissible Human Intervention

AHI-3 is the first level at which the architecture establishes the conditions necessary for the human intervention itself to carry admissible standing.

The route can establish:

Human Standing

Competence

Authority

Scope

Evidence Sufficiency

Evidence Continuity

Verification Feasibility

Operating Condition

Independence

Refusal Capacity

Determination State

Recordability

The system can answer:

Who made the determination?

What role were they performing?

What evidence did they actually receive?

What reality did that evidence represent?

Was it still current?

What authority did they possess?

What scope applied?

Could they meaningfully verify the proposition?

Could they refuse?

What did they determine?

At AHI-3, human intervention moves beyond presence and workflow participation.

It becomes a governed consequence-bearing event.

AHI-3 PRINCIPLE

The human does not merely participate. The intervention has standing.

AHI-4

Pressure-Survivable Human Intervention

AHI-4 asks whether AHI-3 remains true when the system is under real operational pressure.

This is critical.

A human control that works only when:

there is plenty of time,

nothing is failing,

nobody is demanding completion,

evidence is clean,

and rollback is inexpensive

has not yet proven that it can survive the conditions in which governance matters most.

AHI-4 therefore requires evidence that intervention remains operationally meaningful under pressures such as:

time compression,

procedural momentum,

fatigue,

hierarchy,

financial pressure,

dependency accumulation,

rollback cost,

authority drift,

evidence degradation,

boundary saturation,

and partial execution.

The person must still be able to:

HOLD

REFUSE

ESCALATE

CORRECT

and where required:

PREVENT COMMITMENT

before consequence attaches.

AHI-4 PRINCIPLE

Authority that disappears under pressure is not sufficient authority.

AHI-5

Route-Complete Human Governance

AHI-5 is the strongest AHIA level.

It asks whether the entire consequence-bearing human route is governed and replayable.

The architecture can reconstruct the progression from:

REALITY

↓

RECORD

↓

CONTINUITY

↓

HUMAN STANDING

↓

EVIDENCE

↓

DIAGNOSIS

↓

DETERMINATION

↓

AUTHORITY

↓

SCOPE

↓

BINDING

↓

COMMIT

↓

EXECUTION

↓

EXECUTION FIDELITY

↓

OUTCOME

↓

FUTURE RELIANCE

Where human intervention occurs, the system can show exactly where and why.

Where human intervention does not occur, the system can establish why the route was appropriately governed without it.

AHI-5 therefore does not mean:

Maximum human involvement.

It means:

Route-complete governance of human involvement.

AHI-5 Does Not Require a Human Everywhere

This distinction is fundamental.

A highly mature system may deliberately remove humans from certain determination boundaries.

If governed evidence can establish a determination more reliably through:

timestamped records,

multiple corroborating witnesses,

preserved continuity,

declared thresholds,

deterministic criteria,

bounded authority,

and controlled execution,

then adding discretionary human judgment may decrease admissibility.

AHI-5 therefore recognizes both:

admissible human intervention

and

admissible non-intervention.

The architecture should be able to explain why a human was included or excluded.

The Maturity Progression

The six levels can be understood as six increasingly demanding questions.

AHI-0

Is a human present?

↓

AHI-1

Can we prove the human participated?

↓

AHI-2

Is the human's role and authority bounded?

↓

AHI-3

Was the human intervention admissible?

↓

AHI-4

Did that admissibility survive operational pressure?

↓

AHI-5

Can the complete consequence-bearing route be reconstructed and governed?

That progression transforms vague human oversight into something capable of evidence-based evaluation.

AHI Levels Are Not Certifications

AHI levels should not be treated as decorative maturity badges.

A system should not claim:

“We are AHI-5.”

without evidence supporting the relevant consequence routes.

AHI classification should be:

decision-class specific

route specific

evidence based

time bounded

reviewable

A system may demonstrate AHI-4 for one consequence class and only AHI-1 for another.

That is not inconsistency.

It is more truthful governance.

Classification Must Follow Evidence

The architecture should not begin by selecting the desired AHI level.

It should examine the evidence and determine which level the route actually supports.

Evidence → Classification

not

Desired Classification → Supporting Narrative

This preserves the evidence-first posture of TA-14.

AHI Levels Can Degrade

A route that previously supported AHI-4 may later fall to AHI-2.

Why?

Authority may change.

Training may lapse.

Evidence systems may degrade.

Escalation paths may disappear.

Automation volume may overwhelm human capacity.

Verification burdens may increase.

The intervention window may shrink.

Organizational pressure may change.

AHI levels therefore represent a current governance condition, not a permanent status.

The Human-Free Paradox

AHIA produces an important result.

Sometimes a system with less human involvement may possess stronger governance than a system with more human involvement.

Consider:

SYSTEM A

Every determination requires a manager's approval.

Evidence is weak.

Review is rushed.

Bias is possible.

Approvals are routinely rubber-stamped.

SYSTEM B

Reality is captured through timestamped evidence.

Multiple independent witnesses corroborate the condition.

Continuity is preserved.

Declared thresholds govern diagnosis.

Deterministic criteria govern the resulting state.

Humans enter only when ambiguity, exception, authority, rights, or consequence class requires human judgment.

System B has fewer humans in the routine determination path.

Yet it may possess stronger admissibility.

Human quantity is not a governance metric.

The AHI-5 Design Preference

Where a determination can be governed reliably without discretionary human contamination:

Preserve the evidence.

Preserve the sequence.

Preserve continuity.

Apply the governed criteria.

Keep unnecessary discretion out.

Where a human is genuinely required:

Establish Human Standing.

Establish authority.

Establish scope.

Preserve evidence.

Protect independence.

Preserve refusal.

Make intervention survivable.

Bind the determination before consequence.

This is the balance AHIA is designed to establish.

Minimum Evidence Expectations

A practical AHI assessment might examine evidence such as:

AHI-0

Human assignment or availability.

AHI-1

Identity and review timestamp.

AHI-2

Role, authority, scope, and intervention-state evidence.

AHI-3

Human Standing plus evidence, continuity, verification, refusal, and determination evidence.

AHI-4

Pressure testing, intervention-window evidence, escalation survivability, capacity, and fail-closed behavior.

AHI-5

Route-complete evidence connecting the human intervention to binding, commit, execution, outcome, and future reliance.

The higher the claim, the stronger the evidence burden.

AHI Assessment Questions

When evaluating a human governance route, ask:

AHI-0

Is a human present?

AHI-1

Can participation be proven?

AHI-2

Are role, authority, and scope bounded?

AHI-3

Can Human Standing and admissible determination be established?

AHI-4

Can intervention survive realistic pressure?

AHI-5

Can the entire consequence route be reconstructed from reality through outcome?

The first unsupported answer establishes the current evidentiary limit.

Do Not Inflate the Level

A system should not be moved upward merely because it appears sophisticated.

Advanced AI does not establish advanced human governance.

A large organization does not establish strong intervention architecture.

A respected professional does not establish Human Standing for every decision.

A detailed audit log does not establish admissibility.

A policy document does not establish intervention survivability.

Claim only the level the evidence supports.

AHI Levels and Admissible Non-Occurrence

Higher AHI levels should also preserve evidence of consequences that were prevented.

If a human:

held an execution,

denied an unsupported route,

escalated an unresolved condition,

corrected a defect,

or refused commitment,

that event may demonstrate the governance architecture working correctly.

The record should preserve:

what could have happened,

why it did not happen,

what state prevented it,

who possessed standing,

and what happened next.

A prevented consequence can be just as important as an executed one.

AHI Levels and Future Reliance

AHI-5 extends beyond the immediate intervention.

The outcome of today's human determination may become evidence for tomorrow's decision.

Therefore the architecture must ask:

Was the outcome verified?

Did execution remain faithful?

Did the intervention produce unexpected consequence?

Should the determination become precedent?

Should it remain an isolated exception?

Can future systems safely rely upon it?

Human governance should not create unexamined precedent merely because an earlier decision appeared successful.

The AHI Classification Principle

The purpose of AHI-0 through AHI-5 is not to prove that humans are trustworthy.

Nor is it to prove that machines are trustworthy.

It is to determine whether the architecture governing human participation is sufficiently supported for the consequence it is being asked to carry.

The governing progression is:

Presence

Participation

Bounded Intervention

Admissible Intervention

Pressure-Survivable Intervention

Route-Complete Governance

And the highest level does not ask:

How many humans were involved?

It asks:

Can we prove why human intervention occurred, whether it was admissible, whether it remained admissible under pressure, how it affected consequence, and what reality followed?

That is a much stronger standard than:

Human-in-the-loop.

Next: Minimum Human Intervention Evidence Package

AHI levels tell us how strong the human-governance route is.

The next page establishes what evidence must actually exist to prove it.

We will build the Minimum Human Intervention Evidence Package — MHIEP:

Identity & Role

Standing

Competence

Authority

Scope

Evidence Presented

Continuity

Verification Feasibility

Operating Condition

Conflict / Independence

Intervention State

Determination

Binding & Commit

Execution Fidelity

Outcome

Admissible Non-Occurrence

[EXPLORE THE MINIMUM HUMAN INTERVENTION EVIDENCE PACKAGE →]

TA-14 Admissible Human Intervention Architecture

Human quantity is not a governance metric.

Claim only the level the evidence supports.

No admissible evidence. No admissible execution.