---
title: "Authority Finality™ — the category KYE Protocol™ defines and leads · KYE Protocol™"
description: "A correct answer can still be unauthorised. Authority Finality™ is the category KYE Protocol™ defines and leads: proving every consequential AI-agent action was authorised, purpose-bound, evidenced, replayable and final before it changed the world. Output is not authority. Capability is not authority. No authority, no finality."
url: https://kyeprotocol.com/authority-finality/
lang: en
source: "KYE Protocol"
---

> A correct answer can still be unauthorised. Authority Finality™ is the category KYE Protocol™ defines and leads: proving every consequential AI-agent action was authorised, purpose-bound, evidenced, replayable and final before it changed the world. Output is not authority. Capability is not authority. No authority, no finality.

# A correct answer can still be unauthorised. Authority Finality™ settles it.

**Authority Finality™** is the category [KYE Protocol™](https://kyeprotocol.com/) defines and leads: proving every consequential action was **authorised**, purpose-bound, evidenced, replayable and **final** before it changed the world. When an AI agent moves money, releases a clinical result, or files a return, KYE™ proves who authorised it — turning an auditor's days of exam-prep into minutes.

[Start a governed pilot](https://kyeprotocol.com/poc/) [See the stack](https://kyeprotocol.com/agent-authority-stack/)

## The doctrine: no authority, no finality

If you are a CISO or a regulator, every word below is a control you can test. Each rung rules out a thing the market keeps mistaking _for_ authority.

- **Output is not authority.** A model producing an answer is not permission to act on it.
- **Evidence is not authority.** A tamper-evident log proves what happened, not that it was allowed.
- **Identity is not authority.** Knowing who the agent is does not establish what it may do.
- **Access is not authority.** Holding a credential is not the same as being authorised for this act.
- **Capability is not authority.** A benchmark pass proves competence, not mandate.
- **Safety is not authority.** A refusal to do harm is not a mandate to do this.
- **Semantic alignment is not authority.** Understanding the request is not permission to fulfil it.
- **Policy compliance is not authority.** Matching a policy is not the same as being authorised under it.
- **Behavioural conformance is not authority.** Acting as expected is not evidence anyone sanctioned the act.
- **Permission is not mandate.** Being allowed is not being instructed by someone entitled to instruct.
- **Approval is not authority.** A recorded yes proves someone approved, not that they could bind.
- **Technical authority is not institutional authority.** A system granting access cannot grant the institution's consent.
- **No authority, no finality.** An action becomes final only once KYE™ proves it was authorised.
- **Legitimate authority at the exact moment of consequence.** Not before, not after — at the instant the act becomes real.

## The lifecycle KYE™ owns — entity to finality

Adjacent tools each answer one fragment of the question. KYE Protocol™ owns the whole chain a consequential action travels, end to end — eleven stages, each resolving to a control you can audit.

- **Entity** — who or what exists and can act. _Who/what is this?_
- **Principal** — the agent as a first-class governed principal, not an anonymous script. _Is the actor accountable?_
- **Delegated authority** — who authorised whom, for what scope. _On whose mandate?_
- **Purpose Permission™** — authority bound to a purpose, so yesterday's grant cannot authorise today's different act. _For what purpose?_
- **Allowed data, tool, model** — only the inputs this actor is permitted to use. _With what?_
- **Approval** — required human sign-off captured for high-stakes or irreversible acts. _Who approved it?_
- **Decision Map™** — the inputs, policy, evidence and actor that produced the verdict. _How was it decided?_
- **[Evidence Pack™](https://kyeprotocol.com/evidence-pack/)** — a signed, action-level proof bundle. _What is the proof?_
- **Replay-Proof™** — any auditor verifies the seal offline from published keys alone, without trusting KYE™. _Can you prove it independently?_
- **Finality Gate** — the action becomes final only when authority is proven; draft, advisory and executed are distinct states. _May it be relied upon?_
- **Drift & revocation** — authority that expires, drifts or is revoked is caught and itself evidenced. _What happens when a control fails?_

## Everyone else is a point on this map

The category is forming, and serious players are converging on it. That validates Authority Finality™ — and it is why KYE Protocol™ draws the map rather than sitting on someone else's. Each adjacency answers a fragment; KYE™ owns the chain that makes an action final.

| Adjacent player | What it owns | Where it sits on KYE™'s map |
| --- | --- | --- |
| Output-to-action boundary tools | the regulated checkpoint where AI output becomes action | a single gate **inside** KYE™'s chain — KYE™ proves the whole authority that earns the right to reach it |
| Runtime integrity controls | the transition between possibility and consequence at the point of execution | that transition **is** the Finality Gate; KYE™ governs the ten stages that precede it |
| Decision-evidence integrity tools | tamper-evident records of what an AI decided | an **input layer** feeding KYE™'s Evidence Pack™ — KYE™ adds whether the act was authorised |
| Capability benchmarks | whether an agent _can_ do the work | an **input layer**: a benchmark result scopes authority, but capability is not authority |

KYE Protocol™ integrates these layers rather than forking them, ships its own governed agents on top of them, and consumes their signals — the way it already consumes inference-gateway telemetry and capability evaluations — and turns them into authority decisions and replay-verifiable evidence. Govern the action; integrate the input.

Three instruments, one governor. A governance is not one lever: it **enables** (the public goods every agent stands on: identity, vocabulary, runtime, sandboxes, the directory), it **constrains** (purpose scope, allowed actions, obligations, stop conditions, the verdict that makes an action final or refuses it), it **promotes** (certification, sector packs, the programmes that lift selected producers) and it **supports** (replay, auditability, rights). Authority Finality™ is the constraining core; KYE Protocol™ carries all four, and every rail declares which one it is.

## What this means for you

For a board, a CISO, or a regulator, Authority Finality™ converts an unanswerable question — "was this AI action allowed?" — into a signed record you can replay on demand.

- **Regulators and auditors:** every consequential action arrives with its authority basis attached, so a review that took days of reconstruction becomes a minutes-long replay.
- **CISOs and risk owners:** authority can be revoked mid-flight and the revocation is itself evidenced — you control the blast radius of an agent that drifts.
- **Builders:** bring any runtime, policy engine or model; KYE Protocol™ owns the authority lifecycle above it and makes each action regulator-provable across the EU AI Act, NIST AI RMF and ISO/IEC 42001.

[Start a governed pilot](https://kyeprotocol.com/poc/)

Canonical KYE™ surfaces referenced on this page: [Purpose Permission](https://kyeprotocol.com/glossary/).