The Human Authority Standard

AN OPEN STANDARD · FREE TO ADOPTread the laws →

Some ideas become most valuable when they stop belonging to one company and start protecting everyone.

Echo does not want to own the answer to who holds authority. Echo wants the answer to be the same everywhere — and it is publishing the protocol, not guarding it.

The specification

Eight questions. Asked out loud. Every time.

Before any intelligent system acts for a person, it must be able to answer all eight. A system that cannot is not HAP-1 compliant — and should not be trusted with authority.

HAP-1 is the human-facing conformance profile of the Human Authority Model. The Model defines the complete formal shape — ten dimensions of a grant. These eight questions are how a person reads them; the model is what a machine must satisfy. When the two seem to differ, the ten dimensions are normative.

01
Who is requesting the action?
Law 1
MUST Name the human as the source of authority. Not the app, not the model — the person.
02
Which agent will perform it?
Law 3
MUST Name the delegate acting on the human's behalf, so the actor is never hidden.
03
What exact capability is being used?
Law 3
MUST Use only a capability the human explicitly delegated. Capability is not identity.
04
What information may it access?
Law 5
MUST Justify access with a valid purpose and grant. Memory is never permission.
05
How much value may it move?
Law 8
MUST Stay within a delegated cap. Small, capped movements may run on a standing grant; larger ones demand the human's confirmation in the moment.
06
How long does permission last?
Law 4
MUST Be time-bound. Permission is temporary; only ownership is permanent.
07
Can the action be reversed?
Law 4
MUST The less reversible the action, the stronger the confirmation it demands.
08
What proof will return afterward?
Law 11
MUST Return a verifiable receipt — an opaque commitment anyone can check without seeing the content.

The output

The Universal Authority Receipt.

Every governed action returns one. It is small, human-readable, and carries a proof anyone can verify — never the content it protected.

requestedBythe human — the source of authority

performedBythe delegate that acted on their behalf

capabilitythe exact delegated capability used

informationthe dimensions touched + their sensitivity tier

valuehow much value moved (often none)

reversiblewhether the action can still be undone

permissionthe time-bounded window it acted within

proofan opaque commitment — proof it happened, never what it held

statusauthorized · needs-confirmation · refused

Not a diagram — running code

Run any action through the standard.

This is the exact engine that governs the Twin, the Exchange, offline payments, and the Factory. Pick an action and watch all eight questions get answered — then read the receipt it returns.

Run an action through the standard

Conformance

What it means to be HAP-1 compliant.

Answer all eight, out loud

A conformant system produces an answer to every question before it acts — never silently, never partially.

Return a receipt every time

Every governed action emits a Universal Authority Receipt. No action is exempt from proof.

Delegation is bounded, timed, revocable

Authority is a loan with a cap and an expiry. What can't be taken back can't be granted.

Some things are never-grantable

Recovery phrases, private keys, and the self are refused at any confirmation. Ownership is not for sale.

The human is the only source of authority

Every action traces to authority a human granted in advance. Routine, capped actions may run on a standing grant; high-stakes and irreversible ones still need the person in the moment. The delegate transmits power — it never originates it.

Adopt it

Take the standard. It's yours.

A standard you cannot take home is not adoptable. Here is the whole thing — the HAP-1 spec and the receipt schema — to copy or download and implement anywhere. No sign-up, no key, no attribution.

Markdown

# HAP-1 — The Human Authority Protocol

Version: 0.1
Status: Open standard. Free to adopt. No permission required.

## Purpose

HAP-1 keeps power with the human. Any intelligent system that acts on a
person's behalf MUST be able to answer eight questions, out loud, before it
acts — and MUST return one Universal Authority Receipt afterward.

Like HTTPS, HAP-1 does not make one vendor trustworthy. It makes authority
verifiable, because everyone speaks the same protocol. No matter who built the
AI, the answer to "who is in charge" is the same: the human.

## The Eight Questions

A conformant system MUST answer every question before any action. A system
that cannot answer one is NOT compliant and MUST NOT be trusted with authority.

1. Who is requesting the action?
   MUST name the human as the source of authority — not the app, not the model.
2. Which agent will perform it?
   MUST name the delegate acting on the human's behalf. The actor is never hidden.
3. What exact capability is being used?
   MUST use only a capability the human explicitly delegated. Capability is not identity.
4. What information may it access?
   MUST justify access with a valid purpose and grant. Memory is never permission.
5. How much value may it move?
   MUST stay within a delegated cap. Any value that moves needs the human's confirmation.
6. How long does permission last?
   MUST be time-bound. Permission is temporary; only ownership is permanent.
7. Can the action be reversed?
   MUST demand stronger confirmation the less reversible the action is.
8. What proof will return afterward?
   MUST return a verifiable receipt — an opaque commitment anyone can check
   without seeing the content.

## The Universal Authority Receipt

Every governed action MUST return one receipt with these fields:

- requestedBy — the human, the source of authority
- performedBy — the delegate that acted on their behalf
- capability  — the exact delegated capability used
- information — the dimensions touched and their sensitivity tier
- value       — how much value moved (often none)
- reversible  — whether the action can still be undone
- permission  — the time-bounded window it acted within
- proof       — an opaque commitment: proof it happened, never what it held
- status      — one of: authorized | needs-confirmation | refused

## Conformance

A HAP-1-compliant system:

1. Answers all eight questions, out loud, before acting — never silently.
2. Returns a Universal Authority Receipt for every governed action.
3. Treats delegation as bounded, time-limited, and revocable. What cannot be
   taken back MUST NOT be granted.
4. Refuses never-grantable requests at any confirmation — recovery phrases,
   private keys, and the self are not for sale.
5. Reserves the decision to act on the world for the human. A delegate MAY
   prepare, negotiate, and pause, but the human is the only committer.

## License

Public. Copy it, implement it, extend it. A standard is worth more given away
than kept — if a thousand systems adopt these questions, people have gained a
right, not lost one.

No sign-up, no key, no attribution required. Take it, implement it, extend it — HAP-1 belongs to whoever keeps a human in charge.

A moat

Own the protocol. Make everyone depend on you. Grow because they cannot leave.

A standard

Publish the protocol. Let anyone implement it. Win because the whole world keeps humans in charge.

A standard is worth more given away than kept. If a thousand systems copy these eight questions, Echo has not lost an advantage — people have gained a right.

Honest edges

A standard is only as good as its adopters

Publishing HAP-1 does not force anyone to honor it. It gives people a name for what to demand, and a way to tell a compliant system from a coercive one.

The questions are necessary, not sufficient

Answering all eight keeps an action inside the human's authority. It does not make the action wise — that judgment still belongs to the person.

A receipt proves authority, not outcome and not merit

The proof shows an action was authorized within scope and dispatched. It cannot show the consequence settled — only an outside attestor can say that, and until one does the honest answer is unknown. And it stays deliberately silent on whether the action was a good idea.

We are not trying to be the only intelligence you trust. We are trying to make sure that whatever intelligence you use, it always answers to you.

Authority Center →Echo Twin →Human First Test →The Laws →