Saved in:
Bibliographic Details
Main Author: LI, WENXIN
Format: Recurso digital
Language:
Published: Zenodo 2026
Subjects:
Online Access:https://doi.org/10.5281/zenodo.20055225
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866901393980260352
author LI, WENXIN
author_facet LI, WENXIN
contents <p>Abstract<br>The CKS pattern's foundational "human-governed" commitment names three rights — to inspect, to<br>modify, and to override substrate content and orchestration rules — as exercisable at any time during<br>the substrate's existence. The commitment is architectural: the rights must be available as a property of<br>the system's design, not as a procedural promise that depends on a particular vendor, deployment, or<br>workflow (§3.3, §6.3). This note formalizes the canonical failure of that commitment at the<br>AI-infrastructure-vendor boundary. Vendor-revocable governance is the anti-pattern in which the<br>deployment's exercise of the three rights is operationally contingent on the AI infrastructure vendor's<br>continued cooperation: terms-of-service revisions, account suspensions, infrastructure modifications,<br>vendor-side configuration changes, or platform-level policy enforcement can each, at vendor discretion,<br>disable inspection, modification, or override. The rights appear architecturally exercisable but are not.<br>The note specifies the anti-pattern as four operational components, identifies seven CKS commitments<br>it violates, traces the failure mode, names the three-commitment architectural correction (substrate<br>portability per the source paper's tool-agnosticism commitment, authority-structure substrate-residence<br>per the source paper's source-of-truth commitment, governance exercisable through architectural<br>mechanisms the vendor cannot remove), distinguishes the anti-pattern from four adjacent patterns<br>commonly conflated with it, and provides an operational test with three sharpening properties.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_20055225
institution Zenodo
language
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle Vendor-Revocable Governance: An Anti-Pattern Formalization for the Coordination Knowledge Substrate Pattern
LI, WENXIN
Artificial intelligence
Artificial Intelligence
<p>Abstract<br>The CKS pattern's foundational "human-governed" commitment names three rights — to inspect, to<br>modify, and to override substrate content and orchestration rules — as exercisable at any time during<br>the substrate's existence. The commitment is architectural: the rights must be available as a property of<br>the system's design, not as a procedural promise that depends on a particular vendor, deployment, or<br>workflow (§3.3, §6.3). This note formalizes the canonical failure of that commitment at the<br>AI-infrastructure-vendor boundary. Vendor-revocable governance is the anti-pattern in which the<br>deployment's exercise of the three rights is operationally contingent on the AI infrastructure vendor's<br>continued cooperation: terms-of-service revisions, account suspensions, infrastructure modifications,<br>vendor-side configuration changes, or platform-level policy enforcement can each, at vendor discretion,<br>disable inspection, modification, or override. The rights appear architecturally exercisable but are not.<br>The note specifies the anti-pattern as four operational components, identifies seven CKS commitments<br>it violates, traces the failure mode, names the three-commitment architectural correction (substrate<br>portability per the source paper's tool-agnosticism commitment, authority-structure substrate-residence<br>per the source paper's source-of-truth commitment, governance exercisable through architectural<br>mechanisms the vendor cannot remove), distinguishes the anti-pattern from four adjacent patterns<br>commonly conflated with it, and provides an operational test with three sharpening properties.</p>
title Vendor-Revocable Governance: An Anti-Pattern Formalization for the Coordination Knowledge Substrate Pattern
topic Artificial intelligence
Artificial Intelligence
url https://doi.org/10.5281/zenodo.20055225