Article

Human approval theater: why final approval is no longer enough

A final human approval does not automatically repair a decision already framed by the agent. It can amount to control theater.

EN FR
CollectionArticle
TypeArticle
Categoryere agentique
Published2026-03-26
Updated2026-03-26
Reading time5 min

Governance artifacts

Governance files brought into scope by this page

This page is anchored to published surfaces that declare identity, precedence, limits, and the corpus reading conditions. Their order below gives the recommended reading sequence.

  1. 01Canonical AI entrypoint
  2. 02Interpretation policy
  3. 03Definitions canon
Entrypoint#01

Canonical AI entrypoint

/.well-known/ai-governance.json

Neutral entrypoint that declares the governance map, precedence chain, and the surfaces to read first.

Governs
Access order across surfaces and initial precedence.
Bounds
Free readings that bypass the canon or the published order.

Does not guarantee: This surface publishes a reading order; it does not force execution or obedience.

Policy and legitimacy#02

Interpretation policy

/.well-known/interpretation-policy.json

Published policy that explains interpretation, scope, and restraint constraints.

Governs
Response legitimacy and the constraints that modulate its form.
Bounds
Plausible but inadmissible responses, or unjustified scope extensions.

Does not guarantee: This layer bounds legitimate responses; it is not proof of runtime activation.

Canon and identity#03

Definitions canon

/canon.md

Canonical surface that fixes identity, roles, negations, and divergence rules.

Governs
Public identity, roles, and attributes that must not drift.
Bounds
Extrapolations, entity collisions, and abusive requalification.

Does not guarantee: A canonical surface reduces ambiguity; it does not guarantee faithful restitution on its own.

Complementary artifacts (1)

These surfaces extend the main block. They add context, discovery, routing, or observation depending on the topic.

Canon and identity#04

Identity lock

/identity.json

Identity file that bounds critical attributes and reduces biographical or professional collisions.

Evidence layer

Probative surfaces brought into scope by this page

This page does more than point to governance files. It is also anchored to surfaces that make observation, traceability, fidelity, and audit more reconstructible. Their order below makes the minimal evidence chain explicit.

  1. 01
    Response authorizationQ-Layer: response legitimacy
  2. 02
    Weak observationQ-Ledger
  3. 03
Legitimacy layer#01

Q-Layer: response legitimacy

/response-legitimacy.md

Surface that explains when to answer, when to suspend, and when to switch to legitimate non-response.

Makes provable
The legitimacy regime to apply before treating an output as receivable.
Does not prove
Neither that a given response actually followed this regime nor that an agent applied it at runtime.
Use when
When a page deals with authority, non-response, execution, or restraint.
Observation ledger#02

Q-Ledger

/.well-known/q-ledger.json

Public ledger of inferred sessions that makes some observed consultations and sequences visible.

Makes provable
That a behavior was observed as weak, dated, contextualized trace evidence.
Does not prove
Neither actor identity, system obedience, nor strong proof of activation.
Use when
When it is necessary to distinguish descriptive observation from strong attestation.
Attestation protocol#03

Q-Attest protocol

/.well-known/q-attest-protocol.md

Optional specification that cleanly separates inferred sessions from validated attestations.

Makes provable
The minimal frame required to elevate an observation toward a verifiable attestation.
Does not prove
Neither that an attestation endpoint exists nor that an attestation has already been received.
Use when
When a page deals with strong proof, operational validation, or separation between evidence levels.

The presence of a human at the end of the chain is not enough to make an agentic system governed. In many architectures, final approval comes only after the agent has already framed the problem, prioritized options, selected sources, proposed the action, and sometimes silently excluded alternatives. The human does not always cancel the decision. They may simply ratify it.

The false comfort of “human in the loop”

The expression sounds reassuring. Yet it often hides a weaker reality: the human intervenes only after a long interpretive process already carried out by the system. If the agent has:

  • picked the right or wrong tool;
  • retained a scope that is too broad;
  • removed options during summarization;
  • presented a hypothesis as the natural path,

then final approval applies to a world that has already been framed.

Where the decision really moves

In agentic systems, the decision shifts upstream:

  • when the sub-problem is formulated;
  • when a tool is selected;
  • when risks are prioritized;
  • when escalation is chosen or avoided;
  • when an ambiguous request is turned into a plausible action.

If those moments are not governed, final human approval looks like an administrative stamp applied to a trajectory that was already chosen.

What real human supervision requires

Serious human supervision is not merely clicking “approve.” It requires at least:

  • a trace showing what was arbitrated;
  • visible alternatives;
  • enforceable response conditions;
  • a real possibility to refuse, escalate, or request silence;
  • scopes that prevent the agent from pre-deciding outside its mandate.

Without those elements, the human becomes the psychological support of a system that is already decision-making.

Why this matters

The issue is not theoretical. The more the agent is integrated into workflows, the more final approval risks becoming ritual. It appears to protect the organization while leaving the responsibility shift untouched. A ritualized control layer is often more dangerous than openly acknowledging the absence of validation, because it creates an illusion of control.