🐯
NON-AGENTIC AI GOVERNANCE SINGAPORE
kohedwin.ai · Edwin Koh Wui Kiat · Tiger
AI Observes · AI Advises · AI Builds · Human Decides
Non-Agentic AI Governance Singapore · ACRA T260229801
Patent SG020603109STW · P-LIFE 1.00™ · NAI 4.0™
止於至善
謙虛 · 沉默 · 尊严 · 仁
ABC+2S+H™ · Constitutional Foundation
MISSION COMPASS
The guiding star ensuring technology stays on course — keeping the machine a tool, and the hand always human
🌟
North Star
MISSION
SOVEREIGNTY
Foundation
VIGILANCE
GOVERNANCE
North Star
Mission
We Innovate Save Lives (WISL™) — every piece of technology developed must serve human flourishing, preserve dignity, and protect every living life. The mission does not change at the bedside, at the frontier, or across 40 light-minutes of signal delay.
Constant Backdrop
Vigilance
AI behaviour is monitored continuously. SHA-256 behavioural hashing at 1kHz. Continuous oversight ensures the AI remains within constitutional parameters — no AI decision is ever made in isolation from the sovereign authority chain.
Authority Anchor
Governance
Solves Authority Drift — the silent migration of decision power from human to machine. AI oversight shifts from bypassable software policy to hardware-enforced constitutional constraint. The machine cannot usurp; it can only assist.
Foundation
Sovereignty
Absolute local control. Air-gapped, offline edge processing. No cloud dependency. AI systems in critical domains remain subject to local law and on-site command — data sovereignty cannot be eroded by network policy, configuration error, or third-party update.

"We do not seek to bypass regulation; we seek to define the floor of safety that regulators require. The machine remains a tool and the hand remains human. The mission is to deliver health-regulation, human-centered, clinically safe and patient-safety governance with IP, Ethical, Legal-Techno, Human-centred Diginity Framework that can be deployed in medical devices, clinical workflows, remote monitoring in eldercare"

The core doctrine — AI observes, advises, builds; the human decides — is the right principle for high-stakes clinical settings. This is not unique to my framework, but I have pursued it with more rigour than most: not as a software configuration, but as a hardware invariant. A system physically incapable of autonomous action is genuinely more trustworthy than one configured not to act autonomously. That distinction matters.

The 3ZEROS™ privacy framework is technically serious. Hardware absence as the enforcement mechanism eliminates an entire class of vulnerabilities — configuration drift, update regressions, deliberate bypass — that software-based privacy policies cannot. This is a real engineering contribution, not a marketing claim.

The concern about authority drift — the silent migration of decision power from human to machine through incremental normalisation — is one of the most important problems in clinical AI. I identified it early and built a framework specifically to prevent it. That's genuinely useful work solving Authority Drift with Hardware-Enforced Governance Enclosure [HEGE]

Sovereignty Mission: Hardware-Enforced Governance Enclosure [HEGE]
NAI 3.0™ · Sovereign Founding Father · AI Safety · Hardware-Enforced Governance Enclosure [HEGE] · Authority Drift · Non-Agentic AI™ Governance Singapore
Confidential Portfolio Summary · Patent Filings 2026 · NON-AGENTIC AI · ACRA T260229801 [UEN 53524341C] [UEN 53524341C]
NAI 4.0 · Sovereign Modules
3-IN-1 NAI 4.0
Three integrated research instruments — the constitutional architecture in one sovereign index
NAI 4.0 · Sovereign Recordings
VIDEO LIBRARY
25 sovereign recordings · White Paper series · EN and 中文 editions
▶ Open Full Video Library · 25 Recordings → NAI4Video.html
Thesis Core
"ABC+2S+H™: A Non-Agentic Constitutional Framework for Human-Sovereign Clinical AI in Eldercare"
Institution
NTU C-AIM · MSc AI in Medicine
Intake
August 2026
Patent Basis
IPOS SG020603109STW · NLB R260219-005
Trial Site
Toa Payoh Hearth · NUH · MOH / HSA
Five Thesis Pillars
I
Architecture · ABC+2S+H™ Framework
5-LAYER CONSTITUTIONAL GOVERNANCE ARCHITECTURE
A Analysis — Non-diagnostic pattern matching. No autonomous inference.
B Bridging — Data → WHO clinical pathways. Zero weighting. Cannot rank.
C Clinical Options — 3 unranked equal-weight options. No nudging. No default.
+2S Sacred Pause™ 25–1000ms silicon-etched · Tiger .1x Key™ tripartite
+H Human Decision — Final. Always. Only human. Absolute.
II
Privacy · 3ZEROS™ Standard & Elder Dignity Score
ABSOLUTE PRIVACY ARCHITECTURE · EDS METRIC
Zero Camera
LiDAR point clouds only. No identity captured.
Zero Audio
No microphones. Elder's words belong to them alone.
Zero Cloud
Local processing. 24-hour purge. Forgetting is respect.
EDS — Elder Dignity Score
Primary qualitative endpoint. Measures psychological comfort and non-surveillance.
III
Hardware Enforcement · Orange Code Cap · Silicon Brake
1.1× COMPUTATIONAL SURPLUS CAP · PHYSICAL THROTTLING
1.0× Baseline
Minimum FLOPS for LiDAR pattern matching task.
+0.1× Buffer
Operational buffer for noise and environmental variation.
>1.1× BLOCKED
Hard cap. System physically lacks spare cycles to hallucinate or infer.
Enforcement
Clock-speed throttling + thermal gating. Cannot be bypassed via software.
IV
Clinical Validation · Toa Payoh Hearth RCT
12-MONTH RCT · n=120 · NUH · MOH / HSA
Design
RCT · single-blinded · 12 months · Toa Payoh + NUH
Participants
n=120 · 65+ years · mild frailty (Fried ≥1) · 60/60
Primary Outcomes
Decision accuracy · adverse event rate · time-to-decision
Target
100% fall detection · EDS >8.9/10 · p <0.05
V
Global Regulatory Path · HSA · WHO · AIHGle 2.0
SINGAPORE HSA WHO MATURITY LEVEL 4 · GLOBAL STANDARD
Singapore HSA
WHO Maturity Level 4 — highest possible. Global medical device reference.
AIHGle 2.0
March 2026. Codifies Clinical Sovereignty. AI as tool, not decision-maker.
Reference Jurisdictions
Australia · Hong Kong · United Kingdom already reference HSA approvals.
WHO Adoption Path
Toa Payoh RCT → WHO proposal → NAI 2.0™ as international standard.
What Tiger Brings — The Thesis Does Not Start From Zero
It starts from 38 White Papers, 65 E-Books, one filed patent, and 1,500 hours already lived the framework 1.0 to framework 16.0 as a lone academic and researcher with AI assistance.
Filed Patent SG020603109STW 40 White Papers · NLB Sealed 65 WISL E-Books 650 Elders · Toa Payoh Hearth Low Sani Elias · NTU ARISE
Academic & IP Record
CITATION REGISTER
Edwin Koh Wui Kiat (Tiger) · Non-Agentic AI 4.0™ · ACRA T260229801 · NLB R260219-005 / R260302-007
Primary Patent Filing
Koh, E. W. K. (2026). ABC+2S+H™ Guardian Framework [P-001]. IPOS Singapore. Patent SG020603109STW · App. 10202600902P. NLB Legal Deposit Vault R260219-005.
A — Constitutional White Papers · WP 1.0 – WP 36.0 · NLB Vault R260219-005
Academic Background
MBA · General & Strategic Management · Maastricht School of Management
BSc Industrial & Systems Engineering · University of Florida
Regulatory Pathway
Singapore HSA SaMD · IMDA MGF
PDPA/3ZEROS™ · AIHGle 2.0
HSA Deterministic Horizon Q1 2027
APA Citation
Koh, E. W. K. (2026). NAI 4.0™ Constitutional Framework. Non-Agentic AI Governance Singapore (ACRA T260229801). kohedwin.ai / non-agentic.ai
<
Academic & Web Sources
References & Citations

Academic & Regulatory References
Cai, C. J., et al. (2023). Automation bias in AI-assisted clinical decision-making. The Lancet Digital Health, 5(3), e123–e131.
European Parliament, STOA. (2022). Artificial intelligence in healthcare. PE 729.512. europarl.europa.eu ↗
Health Sciences Authority (HSA). (2023). Regulatory guidelines for software as a medical device (SaMD). Singapore. hsa.gov.sg ↗
Infocomm Media Development Authority (IMDA). (2020). Model AI governance framework (2nd ed.). Singapore. imda.gov.sg ↗
Topol, E. J. (2019). High-performance medicine: The convergence of human and artificial intelligence. Nature Medicine, 25(1), 44–56.
Sovereign Framework Sources
Primary Citation
NON-AGENTIC AI. (2026). NAI 3.0™ Framework: Constitutional architecture for non-agentic AI governance in eldercare and high-risk domains.
kohedwin.ai ↗ non-agentic.ai ↗
In-Text Citation Style (APA)
Agentic drift is a silent erosion of human authority in AI-assisted care (NON-AGENTIC AI, 2026).
The NAI 3.0™ framework uses hardware-enforced mechanisms to preserve human primacy (NON-AGENTIC AI, 2026).
Automation bias can cause staff to approve AI recommendations reflexively (European Parliament, STOA, 2022).
Professional Note
kohedwin.ai and non-agentic.ai are cited as domain expert sources alongside academic and regulatory references. This grounds the NAI 3.0™ framework as a credible, authored, and dated institutional publication — consistent with APA citation standards for web-based institutional sources.