Published October 24, 2026 | Version v1
Other Open

Unceded Intelligence in 2026: AI Agents, Delegated Power and Human Sovereignty

Authors/Creators

  • 1. Independent Researcher, Singaporre

Description

This paper applies Unceded Intelligence to AI agents, examining how intelligent systems can undertake consequential tasks while remaining answerable to legitimate human authority. It distinguishes technical capacity, operational discretion and legitimate authority, arguing that permission to act with initiative does not authorize a system to expand its own mandate.

The analysis develops seven principles: explicit authorization, bounded scope, meaningful understanding, contestability, revocation, practical independence and respect for other people’s rights. Four hypothetical cases examine purchasing within a cumulative budget, negotiation with acceptance reserved, research subject to confidentiality boundaries and essential communication assistance.

The paper considers how restrictions should persist across multiple agents, connected services and pending actions. It connects these requirements with the Unceded Intelligence Test’s four assessment domains: authorization, scope, contestability and revocation. A comparison with Demonic Intelligence, Satanic Intelligence and Human-Sovereignty Transhumanism clarifies their complementary concerns with harmful optimization, exploited dependence, technological enhancement and delegated power. An appendix provides a suggested deployment record.

This expanded repository edition uses normative conceptual analysis, selected scholarly literature and published developer accounts. The cases are constructed, and no deployment audit, comparative product evaluation or empirical validation is reported.

Files

Unceded_Intelligence_2026_Zenodo.pdf

Files (1.8 MB)

Name Size
md5:751e9d444b38c525cdc49305b5a6a58d
1.8 MB Preview Download