KTP/
github ↗
created 14 October 2026 · last modified 14 October 2026
may 2026 — the public motion that follows the encyclical

Disarm the Control Plane.

AI does not only need better values. It needs an .

Engagement is not endorsement — the Vatican has not endorsed KTP. Magnifica Humanitas names the moral failure; KTP is one attempt at the engineering surface that answers it.

live · un global dialogue, geneva
The RFP: the UN Panel wrote the requirements. We signed the response.
The campaign thesis in consensus voice: requirements quoted verbatim, the capability map, the signed matrix, red lines drawn red, and a live Geneva log. Open the page →
ACT IThe idea
Alignment
asks what an AI should value.
Authorization
asks who may act, on what, under which constraints, at what cost, and with whose consent.
the locked thesis
Alignment is a category error before it is a safety problem.
Category error → subsidiarity failure → substrate capture.

The bridge: Catholic Social Teaching diagnoses the dignity violation. KTP/TFE explains the control surface that produces it.

where to start
Read this first — the words this page uses (open if unfamiliar)
SubstrateThe relational ground beneath any action — the conditions that allow what an agent (human or AI) can do to become what it actually does. In this campaign: the layer above the model where authorization lives.
AuthorizationThe act of permitting an action. Not the same as having permission. Authorization asks whether an eligible action can be safely absorbed by the environment, right now, given everything else happening.
Control planeAn engineering term for the runtime layer that decides whether a model's output is allowed to become real action in the real world. The campaign argues this is where AI safety has to be enforced — not at the model layer.
A ≤ EPronounced 'A less than or equal to E.' The Zeroth Law of the framework: agent autonomy (A) cannot exceed the environment's trust capacity (E). When violated, the action does not proceed. Stated by Chris in November 2025; signed (in different vocabulary) by Pope Leo XIV in May 2026.
EncyclicalA formal letter from the pope to the bishops and (since the 20th century) to all the faithful, on a matter of universal concern. The highest-frequency form of ordinary papal teaching.
Magnifica HumanitasLatin for 'magnificent humanity.' Pope Leo XIV's first encyclical, signed May 15, 2026 — the exact 135th anniversary of Pope Leo XIII's Rerum Novarum (1891). The first magisterial document on AI in the Catholic tradition.
MagisterialOf or pertaining to the official teaching authority of the Catholic Church. A 'magisterial document' carries the institutional weight of two millennia behind it; an opinion piece does not.
SubsidiarityA Catholic principle (formalized 1931) that decisions should sit at the lowest competent level. Higher levels should help (sub-sidium, 'help from below'), not absorb. The campaign argues Pope Leo XIV re-aimed this principle from state-vs-civic-association to platform-vs-person for the first time.
Category errorGilbert Ryle's 1949 phrase for asking a question whose answer cannot exist in the language of the question. The visitor who tours all the buildings at Oxford and then asks 'but where is the university?' has made a category error. The campaign argues alignment is doing this with AI safety.
RFC / IETF / Internet StandardRFC = 'Request For Comments,' a technical specification document. IETF = the Internet Engineering Task Force, the body that publishes RFCs as Internet Standards. KTP's RFCs are draft specifications — not yet submitted to IETF and not yet Internet Standards. The campaign is honest about this status.
ACT IIThe argument
the three nested claims
Moral claim
No few actors should define the invisible infrastructure of intelligence.
Governance claim
Alignment is a subsidiarity problem before it is a technical problem.
Technical claim
Authorization must be enforced by environmental constraint: A ≤ E.
bridgeKTP turns the moral demand into an engineering surface.
why this is hard to see from inside AI safety
Alignment work is good work. The reason it does not feel insufficient from inside the paradigm is the same reason adding epicycles did not feel insufficient to Ptolemaic astronomy: the instruments point at the model layer, and the model layer is what they can measure. The substrate layer is one step up — closer to authorization than to optimization — and it does not show up in the tools the field already trusts. That is not a failure of intelligence. It is a feature of how paradigms work. Five independent research traditions — biology, physics, complexity science, attractor geometry, and now magisterial — converge on the substrate because they are looking from outside the model-layer streetlight.
Ryle (1949) called it a category error; Kuhn (1962) called it paradigm incommensurability. The two phrases name the same thing from different angles.
the temporal frame

Why authorization has to live in the substrate and not somewhere slower: four clocks run in any system where humans and technology share a world, and only one of them moves fast enough to govern the machine.

Machine
milliseconds
Model inference. Agent action. The Urbina inversion (six hours on a 2015 Mac). The Tuesday 2 a.m. credential-rotation cascade.
Body
years
Trust accumulation between persons. Professional reputation. The slow ripening of a competent practitioner.
Institution
decades
Regulatory adaptation. Doctrinal mutation (¶71 took 95 years to re-aim). Professional norms. RSPs and frontier-lab frameworks.
Aquifer
millennia
Substrate health. Cultural ground. Religious traditions. The relational underlay every faster clock draws from.
The campaign's whole engineering argument: authorization has to run at the machine clock because the model does — and the only way that's possible without surrendering the slower clocks is for the substrate itself to be the gasket between them.
Babel is what happens when one actor controls the substrate others must inhabit. Scattering is the interruption of substrate monopoly.
From the campaign's Babel-as-substrate reading
The Vatican has not endorsed KTP. names a moral failure mode in register; KTP is one attempt at the engineering surface that prevents it. There will be others. The work is to specify the layer in the open, so no single vendor can capture it out of sight.
Why Manning, why now

mediated the 1889 London Dock Strike — entering a frontier labor conflict with moral standing trusted by both sides, and informing Leo XIII through the lead-up to . Olah on the Vatican stage at the Magnifica Humanitas launch is the same institutional move, 137 years later. The Church can host the builder, listen to him, publicly differ from him, and still issue doctrine that constrains the world he is helping build.

ACT IIIThe proof & the honesty

Two tracks

Narrative without proof becomes "AI ethics content." Proof without narrative becomes "yet another access-control spec." The campaign is the bridge between them.

Public narrative

Where the Vatican moment, subsidiarity, disarmament, and platform sovereignty travel.

Flagship essay

The Pope Found the Missing Layer in AI Alignment. Live on Medium, June 12, 2026; the Digital Gravity newsletter followed June 15.

shipped
Magnifica Humanitas — canonical artifact

The encyclical read through KTP. Four engineering-layer gaps the framework supplies.

live
Babel Is a Substrate Problem

Scattering as substrate-level disarmament. The theological warrant for distributed authorization. The standalone essay opens on the Tularosa Basin and closes the loop there.

drafting
From Manning to Olah

The 137-year historical precedent for engagement-is-not-endorsement now lives inside the flagship essay (§7) — Cardinal Manning at the 1889 London Dock Strike as the template for Olah on the Vatican stage.

drafting
The Gödel thread — shipping next

On June 9 NIST put its name behind a proof that no finite guardrail set is complete. "You Can't Patch Your Way to Safe" and "Attribution Is the Prompt Problem Wearing a Trench Coat" carry that proof into the campaign: bound the consequence, not the cause.

drafting

The launch sequence

Seven pieces — flagship first, then the NIST-anchored attribution piece shipping next, then technical, governance, theology, architecture, developer call. The queue is the campaign's transparency. Each row flips to shipped when it lands.

01flagship
The Pope Found the Missing Layer in AI Alignment

May 15 is a Catholic Social Teaching timestamp. Leo placed AI in the labor-capital-platform lineage. KTP supplies the operational answer.

shipped
02technical
Attribution Is the Prompt Problem Wearing a Trench Coat

June 9: NIST proved no finite guardrail set is complete. Attribution inherits the same failure mode — so bound the consequence, not the cause. A ≤ E as the floor that never asks where the action came from.

drafting
03technical
A ≤ E: The Zeroth Law for Agentic Systems

If action risk exceeds environmental trust capacity, the action does not proceed. The Silent Veto, in protocol form.

drafting
04governance
Alignment Is a Category Error Before It Is a Safety Problem

Alignment without distributed authorization becomes substrate capture. The post-Vatican governance move.

drafting
05theology
Babel Is a Substrate Problem

Scattering as substrate-level disarmament. The theological warrant for distributed authorization — the piece that travels outside technical circles.

drafting
06architecture
Blue Zones for AI Agents

A Blue Zone is not a vibe. It is a constrained execution environment where the boundary is enforced by the environment, not asserted by a policy document.

queued
07developer-call
Help Build the Authorization Layer for Agentic Systems

Documentation · RFCs · code · testing · pilots · advocacy. Where builders find their entry point.

queued

Five audiences. Five doors.

Every constituency gets the frame they already recognize and a single best entry point on the site.

Security architects · CISOs
Static IAM will not survive agentic systems. KTP turns authorization into environment-aware constraint.
→ Trust Force Equation
AI safety · alignment
Alignment does not solve value-setting capture. Authorization must be distributed, auditable, and substrate-enforced — and does not require the consciousness question to be settled first.
→ Trust Force Equation
Policy · governance
Regulation alone cannot operate at machine speed. Some constraints must become infrastructure that operators are bound by, not policy operators can revise.
→ Constitution
CST · theology · public ethics
Leo named the dignity violation. KTP explores the operational grammar of disarmament.
→ Magnifica Humanitas read
Open-source builders
Help build the trust proof, Blue Zones, gateway enforcement, telemetry, and audit stack.
→ implement

Honesty about maturity

The architecture is documented. The flagship is shipped. The campaign is in motion. The pilots are not yet built — and naming that is the campaign's first credibility move.
KTP RFC series
draft specification
RFCs (Request For Comments) are the engineering documents Internet protocols are built from; the IETF (Internet Engineering Task Force) is the body that publishes them as Internet Standards. KTP's RFCs are drafts. Not submitted to IETF. Not Internet Standards. Not consensus documents. They are the architecture's working specification.
Constitution + Blue Zones + Trust Force Equation
live on the site
Documented and inspectable. The doctrinal surfaces of the framework.
Magnifica Humanitas artifact
shipped
Canonical reading of the encyclical through the framework. May 26, 2026.
Reference implementations (Python · TypeScript · Go)
under construction
Examples surface partial. Python skeleton in progress. Working pilots: none. Honest pilots before staged pilots.
Campaign essays
flagship shipped · 4 drafting
The flagship shipped to Medium June 12, 2026, with the Digital Gravity newsletter following June 15. The Gödel/attribution thread ships next; Technical, Governance, and Theology drafts sit at v4-FINAL awaiting URLs; Architecture and Developer-Call are queued for after a pilot exists. Status visible in the launch sequence below.
Public engagement on the encyclical's ¶107 from frontier-lab safety leadership
none observed
As of 2026-05-27. Naming the silence is part of the credibility move. If it changes, this row changes.

The corpus this builds on

The Disarm thesis comes out of work the author has been building for almost two years. These are the prior pieces the campaign essays carry forward most directly.

Surviving Adolescence
Feb 2026
The precedent letter to Dario Amodei. Asked Sagan's question of the frontier and named the gap the substrate was supposed to fill.
The Missing Law of Motion
Nov 2025
Where A ≤ E first appeared. The Zeroth Law for autonomous systems, stated as a physical constraint rather than a policy aspiration.
Softened Ground
May 2026
The substrate-decay essay. Four Clocks. The three gaskets that used to absorb the speed differential between systems. What conspiracy theories are really pointing at.
Sailing by Starlight
Dec 2025
The wayfinder cosmology. From administrator to wayfinder. Trust as mass. The closing register of the flagship essay echoes this piece.
Authority Without Gravity
May 2026
Agents of Chaos read through the framework. The engineering-register version of the autonomy-vs-substrate problem the Disarm campaign extends.

Source artifacts

The next layer of AI safety is not alignment. It is authorization: no intelligence should become action without a substrate capable of bearing, auditing, and constraining it.
How to cite
APA 7th ed.

Perkins, C. (2026, May 26). Disarm the Control Plane. Kinetic Trust Protocol. https://kinetic-trust-protocol.net/disarm
Plain text

Chris Perkins, "Disarm the Control Plane," Kinetic Trust Protocol, https://kinetic-trust-protocol.net/disarm.